Stable Release 1.140

Discover how VARIOS AI enables you to continuously expand your AI environment – securely, flexibly, and transparently.

Our magazine article provides an overview of all the latest features – from the secure processing of sensitive documents to the integration of cutting-edge AI models. Every new feature gives you greater control and opens up new possibilities for using AI on your own terms.

Processing text in documents

Connecting systems to VARIOS AI

Document processing with an embedding model

Nano Banana integration

Processing anonymized text from documents

With our integrated Data Loss Prevention (DLP), we check not only chat content but also documents for sensitive data. Only after careful anonymization do we provide the texts to the AI model (LLM).
This lets you use confidential content safely — even in highly regulated environments.

The key advantage

The LLM works on the anonymized texts, not on the complete, potentially personal original documents. This makes it possible to implement AI-driven scenarios in strictly regulated areas without compromising data protection and compliance.

Connect systems to VARIOS AI independently

With OpenAPI and MCP support, you can integrate your systems into VARIOS AI on your own — no presets or external assistance required. You choose which internal and external data sources to use and expand your assistants’ knowledge base step by step.
This increases the reliability of responses and unlocks new use cases. OpenAPI and MCP are the foundation for a custom, growing AI ecosystem — independent, modular, and at your own pace.

Document processing with a configured embedding model

VARIOS AI now lets you process documents directly using your configured embedding model. This gives AI models access to document content that previously could not work with files.
Whether you run the models in Azure, your private cloud, or on-premises, all processing — from vectorization to retrieval – is performed in a data-protection-compliant way within your infrastructure.
Security that enables innovation.

Integration of Nano Banana in VARIOS AI

VARIOS AI now supports the latest Google model generation with an extended focus on Gemini 3 Pro Image. A specially developed connector integrates the image capabilities — image analysis and image generation — directly into VARIOS AI.

Conclusion

The latest update places a clear emphasis on data sovereignty, flexibility, and modernization.

By decoupling document processing and offering the option to use local embedding models, the risk of data leakage is minimized. In addition, the new processing of anonymized documents enables secure analysis of even sensitive content, opening up entirely new application areas.

At the same time, organizations gain maximum independence through the opening of the OpenAPI and MCP interfaces: external systems and data sources can now be connected independently and without support effort. The release is rounded off by technical support for Google Gemini 3 Pro, including the new image AI, ensuring access to the latest model generations.