We're excited to announce the release of metaphactory 5.7 which comes with enhanced vocabulary governance features, including term-based provenance tracking, editorial states and review comments to streamline collaboration, as well as usability improvements to the graph canvas. As part of this release, we are also introducing new capabilities for capturing and integrating schema metadata from physical data sources, as well as foundational enhancements for our AI capabilities.

Term-level provenance & collaboration features for vocabularies

Building on the editorial workflows for vocabularies introduced in the previous release, the key enhancements in metaphactory 5.7 reinforce our commitment to making knowledge graph management more auditable, efficient and collaborative. These enhancements to vocabulary governance include:

• Term-based provenance tracking aligned with the W3C PROV-O standard, allowing users to visually compare changes at different granularities, from the overall vocabulary level down to individual terms.
• Term-based editorial states that help teams organize their workflow by tracking terms through different stages – development, ready for review, in review, accepted for publication – ensuring a more structured approval process. Advanced filtering capabilities help curators to focus on the relevant editorial stages.
• Term-based review comments and decisions that now enable more efficient review cycles, making it easier to discuss and finalize changes directly within the platform.

As a sneak peek into future developments, metaphactory 5.7 also introduces an experimental vocabulary mapping editor, laying the groundwork for advanced alignment capabilities across vocabularies.

Laying the foundation for bridging the gap between physical data sources and ontologies

As part of this release, we are excited to introduce new capabilities for capturing and integrating schema metadata from physical data sources.

To bridge the gap between raw data and semantic models, metaphacts has developed and published a set of meta-models represented as ontologies that describe how to capture metadata from relational databases (e.g., tables, columns, keys, constraints) and JSON documents (e.g., objects, properties, constraints). This lays the foundation for future mappings between the low-level structure of physical data sources and higher-level ontologies, ultimately enabling a seamless connection between raw data and logical models.

With this release, a new metaphactory app is available, which provides an early-stage implementation for converting metadata about relational databases and document schemas into RDF based on these new ontologies. This means users can already start integrating and exploring schema information from physical data sources within their knowledge graphs today.

Usability enhancements to the graph canvas

The new release also brings a series of usability improvements to metaphactory's graph canvas, a core component used across the platform for visual knowledge graph exploration, diagramming and ontology modeling:

• Optimized node design for better readability, maximizing space for labels.
• Improved navigation with state persistence when navigating back and forth.
• Keyboard shortcuts for faster, more seamless interactions.
• Improved group selection which also supports better organization of graph elements.
• Refined styling aligned with metaphactory's design system for a more polished experience.

These updates further enhance usability, making knowledge graph exploration and modeling more seamless, efficient and visually intuitive.

Strengthening AI foundations

With this release, metaphacts additionally strengthens the foundations of its AI capabilities, paving the way for a general availability (GA) release of the conversational AI agent. This release introduces a new backend service management infrastructure that enables the configuration and management of various AI services and tools across the platform.

A key feature is the ability to configure and connect to different LLM providers such as OpenAI, Ollama, Mistral, and Gemini. As part of this enhancement, we've implemented a structured evaluation and testing approach to compare different models in the context of metaphactory’s conversational AI.

Ultimately, these improvements provide the flexibility to utilize different LLM models, either self-hosted or under user control, allowing for optimization across privacy, cost and control. It also opens the door to specialized models being used in parallel for different tools and capabilities within metaphactory's AI agent.

Have a look at the release notes or This email address is being protected from spambots. You need JavaScript enabled to view it. to learn more about this release.

Get started with metaphactory for free »