> ## Documentation Index
> Fetch the complete documentation index at: https://docs.varios-ai.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Global Assistants

> Manage global assistants for your VARIOS AI instance

<Tip>
  Home > "Admin Menu" > "Global Assistants"
</Tip>

## Global Assistants

Global assistants are the heart of VARIOS AI. They provide the actual AI tools that end users will work with later—from simple chatbots to complex, enterprise-specific expert assistants.

A global assistant consolidates all relevant settings and resources into a central configuration. This allows administrators to precisely control how an assistant operates, which data it can access, and which user groups are allowed to use it.

An assistant typically consists of the following components:

* **System Prompt**: defines how the assistant behaves and what level of response quality is expected.
* **AI Models**: selection from various connected language models (e.g., OpenAI, Azure, Claude, Mistral, Llama).
* **Connectors**: direct integrations with external systems and data sources (e.g., Microsoft 365, Bing, internal APIs).
* **Functions and Files**: optional additional features such as Code Interpreter or document processing, as well as uploading your own files.
* **Knowledge Bases**: integration of company-specific documents based on the vector database (RAG).
* **Data Protection**: activation of mechanisms such as GDPR Protection or other DLP rules.
* **Groups**: control of visibility—assistants can be specifically assigned to certain user groups.
* **Description and Examples**: A short description of the assistant, examples of user inputs, and a start heading that is displayed on the home screen.
* **Overrides** (Expert view): Configuration of model-specific settings

Thanks to this modular architecture, assistants can be flexibly adapted to the requirements of individual departments or use cases.\
This creates a unified platform where companies can provide their own AI assistants with clearly defined functionality, vetted data sources, and binding security policies.

<Tip>
  **Best Practice**: Create multiple specialized assistants instead of a universal "all-round assistant." This keeps the scope manageable, the answers more precise, and allows security and data protection policies to be implemented specifically depending on the use case.
</Tip>

<Note>
  **Personal Assistants**: Users can also create their own assistants if they have been granted permission for at least one model in the group settings (connectors optionally as well). The assistant configuration for users, given the appropriate permission, can be found under Home > All Assistants > New Assistant
</Note>

## Configure Global Assistants

<Steps>
  <Step title="Assign a name and image">
    Give the assistant a unique name and optionally upload an image.\
    This makes it easier to distinguish it later in the user interface.
  </Step>

  <Step title="Define the system prompt">
    Define the desired behavior of the assistant in the **system prompt**.\
    Specify which role the assistant should assume, what tone it should use, and how it should handle uncertain inputs.\
    Example: “You are a subject-matter assistant for legal questions within the company.”\
    Design your system prompt individually, revise it with the prompt generator, or select a suitable template from the prompt library.
  </Step>

  <Step title="Select an AI model">
    Choose the language model that should be used for the assistant.\
    VARIOS AI supports various providers and models (e.g., OpenAI, Azure, Claude, Mistral, Llama).\
    The choice of model affects response quality, costs, and data protection.
  </Step>

  <Step title="Add connectors">
    Integrate external systems and data sources via connectors.\
    Examples: Microsoft 365, internal wikis, search services, or security systems.\
    Through connectors, the assistant can access information and tools in real time.
  </Step>

  <Step title="Enable Functions and Files">
    Determine whether files may be uploaded in the assistant and how they should be processed (Expert view):

    * **Local Document Processing** (analysis of uploaded files using the stored embedding model)
    * **Document Processing via the OpenAI API** (analysis of uploaded files via the OpenAI API)
    * **OpenAI Code Interpreter** (execution of code snippets)

    Optionally, files can also be uploaded here that should always be available for conversation in this assistant.
  </Step>

  <Step title="Integrate knowledge bases">
    Select one or more previously created knowledge bases.\
    These serve as the foundation for Retrieval-Augmented Generation (RAG) and enable context-specific answers based on internal documents.
  </Step>

  <Step title="Overrides (Expert View)">
    Override model-specific settings such as creativity (temperature), word diversity (top p), reasoning effort, and output token limit.
  </Step>

  <Step title="Configure data protection">
    Define how personal data should be handled:

    * **Enabled:** All detected data is automatically anonymized.
    * **Optional:** The user is warned and can decide.
    * **Disabled:** Content is not checked.

    Optionally, custom **deny lists** (regex) and **allow lists** (words) can be maintained in the admin portal.
  </Step>

  <Step title="Assign groups">
    Determine which user groups will have access to the assistant.\
    For example, HR, IT, or Sales can use their own assistants tailored to them.
  </Step>

  <Step title="Add description and examples">
    Add a short **description** of the assistant as well as examples of typical user inputs and override\
    the start heading. This makes it easier for end users to get started and promotes consistent usage.
  </Step>

  <Step title="Create and test the assistant">
    Save the configuration and test the assistant directly in the administration interface.\
    Adjust the prompt, data sources, or security settings as needed until the desired behavior is achieved.
  </Step>
</Steps>