Release Notes 22.05.2026 - v3.29.0

📝 This release introduces a new central permissions service — Auth Center — moves user, department, role, and per-feature access into a single place, and the user-management UI has been rebuilt on top of it. Context-level permissions arrive as first-class citizens, a chat-only access role lets organisations expose Minerva without the rest of the tool, and a combined policy decision algorithm stacks user- and department-level roles consistently. Minerva picks up analyzer-aware querying and inline Mermaid rendering, the Builder gains a side-by-side timestamp transformation preview, and the Minerva thinking panel now streams live planner reasoning. Several Minerva, importer, and layout issues are also resolved.

📅 22.05.2026
🔖 Version 3.29.0


✨ Features

  • Central permissions service (Auth Center) 
    A new central permissions service manages users, departments, roles, and per-feature access across the platform, enabling fine-grained permissions throughout the Noreja tool.
  • Rebuilt user-management UI 
    The user-management UI has been rebuilt on top of the new Auth Center. Customer admins can edit users, soft-delete them, resend invitation mails, and trigger password resets, and every user can edit their own profile (including language preference).
  • Chat-only access role 
    Customer admins can grant employees a chat-only access role. Users with only the chat feature are taken directly to the private Minerva chat and do not see the rest of the analytics tool; the role also controls who can create, delete, read, and update chats in the private, company, and detail-chat contexts.
  • First-class context permissions 
    Each company, dimension, and entity-graph context now has its own access list. Admins decide which departments and roles can view or edit a context, and Minerva only uses contexts the current user is allowed to see.
  • Combined permission policy decisions 
    Permissions are evaluated through a single policy decision algorithm that combines user-specific roles with department-assigned roles and applies the highest level of access. A user inherits everything granted by the departments they belong to, and their direct role assignments stack on top.
  • Minerva respects the analyzer view 
    Temporal layer, semantic path, and negative-time-handling settings from the current analyzer tab are applied to Minerva's graph queries, so the AI's analysis matches what is on screen.
  • Inline Mermaid diagrams in Minerva 
    Minerva can now render Mermaid diagrams inline, including a fullscreen view, so process sketches and data visualisations appear directly in the response.
  • Live planner reasoning in the chat thinking panel 
    The Minerva chat thinking panel now shows each planned step with the planner's reason, and on supported models (Anthropic, OpenAI Responses API, on-prem Gemma 4) the model's reasoning tokens stream live above the plan. The raw planner JSON no longer leaks into the UI, and models without a reasoning channel hide the panel rather than show an empty one.
  • Side-by-side timestamp transformation preview 
    The Builder's timestamp transformation editor provides an advanced way to test the transformation with a side-by-side comparison view before the import is triggered.
  • Entity-graph export includes table properties 
    The entity-graph export file now includes database table properties and their transformations, so configurations round-trip cleanly through export, import, and copy.
  • Redesigned analyzer settings panel 
    The analyzer settings panel has a tighter layout, toggle-button grouping for case model / case instance, and a grid arrangement for analyzer views.
  • More reliable process graph layout 
    The dagre layout used to draw process graphs was reworked to handle smaller screens, segmented edges, and rework patterns more reliably.

🐛 Bug Fixes

  • Private chat scoped to its creator 
    The private chat is now strictly scoped to its creator. Attempting to read another user's private conversation returns a 403, closing a potential cross-user data-exposure path.
  • Private chat no longer suggests process follow-ups 
    Asking a simple question in the private chat no longer triggers process-focused follow-up suggestions that the private chat cannot answer.
  • Private chat synthesis resolves correctly 
    The synthesis step in the private chat is no longer left unresolved; the response synthesizer dedicated to private chat is now used.
  • Correct dashboard vs. analyzer context detection 
    Minerva now correctly distinguishes whether you are on the dashboard or in the analyzer view and no longer claims an analyzer-view context when you are on the dashboard.
  • Violations on case start/end activities reach the filter 
    The importer now creates a violation filter entry when a violation occurs on the start or end activity of a case.