Skip to main content

July 2026

Sophia Rahbauer
Sophia Rahbauer
Product Communication @ ADITO

With July 2026, the ADITO platform takes a significant step towards AI-powered business applications. Vector Search enables AI systems to securely access and utilize business data stored in ADITO while respecting existing permissions and access controls.

At the same time, the expanded ADITO AI Runtime introduces speech recognition, semantic embeddings, and an improved language model, providing the building blocks for developing AI-powered applications.

Vector Search for AI-Powered Business Applications​

2026.1.0 core

Vector search provides the foundation for a new generation of AI-powered applications within ADITO. By making selected business data available through semantic search, AI features can understand, retrieve, and use relevant business data directly from the ADITO platform.

Unlike traditional keyword-based search, vector search enables AI systems to find information based on meaning and context. This allows customizers to build intelligent assistants and automation scenarios that work with business data while maintaining full control over data access and permissions.

Vector Search Figure: Selected ADITO data is vectorized and made available for AI-powered applications

  • Make Business Data Available to AI
    AI-powered applications can leverage relevant information stored in ADITO, enabling context-aware assistance and automation.

  • Full Control over Accessible Data
    Customizers decide exactly which entities, index groups, and fields are available for AI use cases.

  • Permission-Aware Data Access
    Existing ADITO permission concepts remain in effect, ensuring users and AI services only access data they are authorized to see.

  • Privacy and Compliance by Design
    Combined with ADITO's hosting and deployment options, vector search enables AI solutions that meet demanding data protection, compliance, and governance requirements.

  • Foundation for AI-Powered Business Applications
    Vector search provides the technological basis for custom AI assistants, intelligent search, knowledge retrieval, and other AI-driven business applications.

To make use of this capability, vector search must be enabled and configured for the relevant index groups and data fields.

ADITO AI Runtime: Expanded AI Capabilities​

The ADITO AI Runtime has been expanded with new capabilities for building AI-powered business applications. Customizers can now integrate speech recognition, semantic embeddings, and an improved language model into their applications through centrally managed AI services, without operating their own AI infrastructure.

Adito AI Runtime Figure: The ADITO AI Runtime provides a unified API for integrating AI capabilities into custom applications, workflows, and processes.

  • Speech-to-Text
    Convert spoken content into text for use in custom applications and workflows. Typical scenarios include meeting notes, visit reports, dictation, and service documentation.

  • Embeddings for Semantic Search and RAG
    Generate vector embeddings to enable semantic search, similarity matching, and Retrieval-Augmented Generation (RAG). Combined with Vector Search, embeddings allow AI applications to retrieve relevant business information based on meaning rather than exact keywords. See the Vector Search via Solr guide for implementation details.

  • Updated adito-llm
    The latest adito-llm generation delivers improved instruction following, more reliable structured outputs, and better support for complex AI workflows while continuing to use the existing integration. According to the official Qwen benchmarks, the latest model generation delivers measurable improvements across coding, reasoning, and agentic workflow evaluations, providing a strong reason to update existing AI-powered applications.

  • Centrally Managed AI Services
    All capabilities are provided through the ADITO AI Runtime. Customizers can focus on building AI-powered business applications while the platform centrally manages the required AI models and infrastructure.

Speech recognition quality depends on recording conditions and language. Semantic search quality depends heavily on the selection, preparation, and chunking of the underlying data. Existing prompts, structured outputs, and AI workflows should be reviewed and tested after upgrading to the latest adito-llm model generation.

For more information, see the Speech-to-Text, Embeddings and Text Generation documentation.

Further Improvements​

  • Variable Graph V2 replaces V1 and provides more reliable dependency tracking across processes and shared libraries. 2026.1.0 Core
  • EntityModel introduces a unified row recalculation model across the user interface, Web Services, and entities.getRows, providing more consistent results and improved performance. 2026.1.0 Core
  • IndexSearch delivers more secure and scalable searches with permission-aware server-side execution and improved support for large datasets. 2026.1.0 Core
  • A dedicated JWT keystore enables secure service-to-service authentication for future platform components outside the Vaadin session. 2026.0.1 Core