Minerva FAQ
Short answers to the questions about Minerva that come up most often. Each answer links to the page that explains it in depth.
Why Can't I See Minerva?
Minerva needs a role with the feature Chat: Read opens the chat page from the profile menu, where you can ask questions; Create adds the Minerva-AI switch next to dashboards and the Analyzer. Ask your administrator. If you see the chat but no agent or no model to choose, no agent is shared with you or no model is available. See Create and Edit Roles and Navigating Noreja.
Why Does Noreja Open Directly in the Chat?
If your role only allows chatting with Minerva, Noreja opens on the chat page without the top bar. On a phone or a small tablet, Noreja opens there for everyone who may use the chat. See Chat with Minerva and Navigating Noreja.
Can Other Users or Admins See My Chats?
Your chats are listed only for you: other users, including Customer admins, don't see your chats in their chat list. Workspaces can be shared, chats can't. See Minerva: Models, Data Flow and Data Protection.
Which AI Model Does Minerva Use?
The one shown under Model below the input field. A chat starts with the agent's default model or the default of your installation, and you can switch at any time. Local models run on servers operated by Noreja or your company; External Models come from Anthropic, OpenAI or Mistral. See Minerva: Models, Data Flow and Data Protection.
What Happened to the Model I Used Before?
When Noreja replaces a model with a newer one, chats that used it continue on the successor automatically, also when you regenerate an answer. The chat doesn't show a notice. Agents whose default model was replaced use the successor too.
Is My Data Sent to an External AI Provider?
Only if the chat runs on an external model. Then your question, the earlier messages of the chat and the data Minerva reads for it go to that provider, including passages Minerva finds in Context Hub documents. Answers you locked and knowledge marked Local LLM are left out. Dictation always uses a speech service from OpenAI. See Minerva: Models, Data Flow and Data Protection.
Why Doesn't Minerva Use My Document?
- The document isn't shared with you. Minerva only uses knowledge you may read.
- It's marked Local LLM and the chat runs on an external model. Switch to a local model.
- It isn't processed yet: wait until it shows Indexed. Knowledge about a perspective only works after the perspective has been imported, knowledge about an activity after its entity has been imported.
- Very short texts are found less reliably. Workspace files aren't part of this search: attach the file or name it in your question.
See Context Hub: Knowledge for Minerva.
Does My File Belong in the Workspace or the Context Hub?
In the workspace if Minerva should work with the file now, for example read, analyse or change a supplier list or meeting minutes: upload it and attach it to your question. In the Context Hub if it's knowledge Minerva should know for many questions, also your colleagues' questions, for example a policy or a process description: Minerva then searches it by itself. Files can't be added to the chat directly. See Work with Files in Minerva Workspaces and Context Hub: Knowledge for Minerva.
Why Isn't My Skill Used?
Minerva loads a skill when your question matches its short description, so describe the task there in the words users ask in. The agent must allow the skill under Allowed skills. To use it for sure, type / and choose it at the start of your message. See Write Your Own Skill.
Can Minerva Change My Data?
No. Minerva only reads process data, perspectives, analyzer tabs and filters; it can't import data, change perspectives, tabs or filters, or build dashboards. It only creates answers and files in workspaces where you may edit. See What Minerva Can Do.
Why Did Minerva Stop Before Answering?
An answer may take a limited number of steps (by default 40 for Minerva and 12 for Minerva (On-Premise); admins can change it per agent under Maximum iterations). Very large tasks can reach the limit. Split the task into smaller questions, or ask again more precisely. See AI Hub: Agents, Skills, Prompts and Workspaces.
How Long Can a Chat Get?
Each model can take a certain amount of text at once, much less for the local model than for external ones. When a chat gets long, Minerva summarises earlier messages automatically; the ring next to the input field shows how full the chat is, and you can summarise manually. Locked answers are never summarised. For a new topic, start a new chat. See Chat with Minerva.
Which Languages Does Minerva Understand?
Minerva answers in the language of your question; German and English are tested. Dictation recognises German and English. Other languages depend on the model and aren't guaranteed.
Why Do I See Token Costs?
Below each answer and under the token usage, Noreja shows how many tokens the models used and an estimated cost. Local models have no price per token. Premium models cost more; Noreja points this out when you choose one. See What a Model Costs.