AI Assistant (Beta)

With the AI Assistant LIA (LOGINventory Intelligent Assistant), you can chat with LOGINventory to get information, ask questions, or give instructions. LIA uses OpenAI's AI technology to understand your requests and provide relevant answers.

Accessing the AI Assistant

You can access the AI Assistant LIA in several ways:

1. Via the quick access bar or the ribbon menu in the LOGINventory user interface.
2. Via the web browser: First, the Web Viewer must be published, and the corresponding URL (e.g., https://my-loginventory/LOGINventory9/lia.aspx) must be opened.
3. When creating queries: Instead of manually selecting which columns to display and which filter conditions to set, you can describe textually what you want to see.
4. Via your own AI that accesses the MCP server of LOGINventory. We are happy to support you in connecting your AI solution—contact our support for more information.

Access to LIA is only available to users with the role "Administrator" (Exception: Query creation also available for "power users"), and an active internet connection is required for LIA to use OpenAI's AI services.

The Chat Interface

The interface is identical in both desktop and web versions and consists of two areas:

• Chat area (left): Here you ask your questions and receive answers from LIA.
• Results area (right): Here, query results are displayed as an interactive data table. This area appears automatically as soon as LIA executes a query.

A movable divider between the two areas allows you to adjust the layout.

Sending Messages

Enter your question or instruction in the text field at the bottom and click the Send button (arrow icon) or press Enter. To insert a line break within the message, use Shift+Enter.

While LIA is responding, you can stop processing using the Cancel button (X icon).

New Chat

Use the "New Chat" option in the toolbar to start a new chat. This resets the chat history and the results table. LIA starts without knowledge of previous messages.

Results Table

When LIA executes a query, the results appear as a table in the right area:

• Sort: Click a column header to sort ascending or descending.
• Filter: Below the column headers is a filter row. Enter a search term there to narrow down the results.
• Record count: The number of results is displayed in the toolbar (e.g., “42 records”).

Features and Capabilities

LIA can perform the following tasks for you:

Create and Execute Queries

Describe in natural language what data you need. LIA automatically creates a query and displays the results in the data table.

Examples:

• “Show me all Windows 11 devices with less than 8 GB RAM”
• “Which devices have Microsoft Office installed?”
• “List all users who have not logged in for 3 months”

You can ask follow-up questions that refer to previous results:

• “Of those, only the ones with SSD”
• “Sort by last login”

Save Queries

LIA can permanently save a created query in the "My Queries" folder:

• “Save this query as ‘Old Windows Devices’”

The saved query is then available in the Management Center and Web Viewer.

Manage Custom Properties

LIA can create custom properties and set values:

• Create property: “Create a custom property ‘Cost Center’ of type text for devices”
• Set value: “Set the cost center of PC-IT-007 to ‘IT-2024’”
• Rename property: “Rename the property ‘Location’ to ‘Building’”

Supported data types: Text, Yes/No, Decimal, Date. The data type cannot be changed after creation.

Create Lifecycle Entries

You can use LIA to create lifecycle entries for devices:

• “Create a repair entry for PC-IT-007 from 03/15 to 03/22”
• “Add a loan for LAPTOP-POOL-03, 4 weeks from today”

LIA understands natural date expressions like “from tomorrow”, “next week”, or “4 weeks from today”.

Asset Actions

LIA can manage individual assets:

• Archive: “Archive the computer PC-HR-042”
• Reactivate: “Reactivate the device LAPTOP-SALES-01”
• Delete: “Delete the peripheral device Monitor-003”

Before each action, LIA shows a preview and asks for your confirmation. Mass actions (e.g., “Delete all old devices”) are not supported for security reasons.

Search the Manual

LIA can search the LOGINventory manual and summarize relevant information for you:

• “How do I scan an Exchange Server?”
• “What is the difference between archiving and deleting?”
• “Explain role management to me”

MCP Server

The MCP Server (Model Context Protocol) allows you to connect an external AI (e.g., Claude, ChatGPT, or your own solution) with LOGINventory. The AI can then use the same tools as LIA—create queries, manage assets, search the manual, and more.

Endpoints

The MCP Server is accessible via the Web Viewer:

Available Tools

The following tools are available via the MCP Server:

Authentication

Access to the MCP endpoints is exclusively via API keys, which are configured in the settings. Access with username/password is not possible. Instead, each API request must include the corresponding API key in the HTTP header X-Api-Key.

Example: Retrieve Tools

GET https://my-loginventory/LOGINventory9/api/mcp/describe
X-Api-Key: <your-API-key>

Response:

{
  "name": "LOGINventory MCP Bridge",
  "version": "1.0",
  "tools": [ ... ]
}

Example: Execute Tool

POST https://my-loginventory/LOGINventory9/api/mcp/execute
X-Api-Key: <your-writer-API-key>
Content-Type: application/json

{
  "toolName": "plain_text_query",
  "argumentsJson": "{\"prompt\": \"Show me Windows 10 computers\"}"
}

Response:

{
  "Name": "Windows 10 Devices",
  "BaseEntity": "Device",
  "Filter": "Contains([OperatingSystem.OperatingSystemInfo.Name], 'Windows') AND Contains([OperatingSystem.OperatingSystemInfo.DisplayVersion], '10')",
  "Properties": [
  { "FullName": "Name", "Selected": true },
  { "FullName": "OperatingSystem.OperatingSystemInfo.Name", "Selected": true }
  ]
}

Data Protection and Security

As mentioned, LIA is only available to users with the role "Administrator" (Exception: Query creation also available for "power users").

The following list describes exactly what happens to your questions when using LIA and what access the AI service has to your data:

1. New chat: Formulate a question or instruction in natural language. E.g., "Show me all devices owned by Max Mustermann".
2. Transmitting the question: Your message is sent unchanged to the AI service. In this example, the message contains the name "Max Mustermann".
3. Processing by AI: The AI service processes the question and generates a query to obtain the requested information. E.g., "Select Name, OperatingSystem.Name, LastInventory.Ip where Device.Owner.Fullname = 'Max Mustermann'".
4. Query results: The query is executed locally in LOGINventory, and the results are displayed in the results table. In this example, the result could be a list of devices owned by Max Mustermann.
5. Up to 25 pseudonymized rows of the query results are now sent to the AI service. In this example, the actual device names are replaced by tokens like ENTITY_000, ENTITY_001 (see Automatic Anonymization) before being sent to the AI service.
6. AI response: The AI service generates a response for the chat conversation based on the question and the anonymized results. In this example, the answer might be: "Max Mustermann owns 3 devices, including ENTITY_000 with operating system ENTITY_001".
7. Displaying the answer: LIA's answer is displayed in the chat area. All anonymized tokens in the answer are automatically replaced with the real values before being shown to you. In this example, ENTITY_000 would be replaced by the actual device name and ENTITY_001 by the actual operating system.

Automatic Anonymization

LIA uses an external AI service (Azure OpenAI) to generate a textual answer to your question. All sensitive inventory data is automatically anonymized before being sent to the cloud service.

Anonymization works as follows:

• All text values in the query results (computer names, user names, locations, etc.) are replaced by anonymous tokens like ENTITY_000, ENTITY_001, etc.
• Numeric values, date values, and GUIDs are not anonymized.
• In your follow-up questions, words that appear exactly as shown in the results table are also automatically replaced by the corresponding token before the message is sent to the AI service.

The AI works exclusively with the anonymized tokens. When displayed to you, the tokens are automatically replaced with the real values—so you always see the actual data.

Example: What is sent?

Suppose you ask: “Show me Windows 10 computers”

LIA internally creates a query and receives results. Before the data is sent to the AI service, anonymization takes place:

What LIA shows you (real data):

What is sent to the AI service (anonymized):

Note: Identical values receive the same token (here ENTITY_001 for “Microsoft Windows 10 Enterprise”). The cloud service cannot infer your real data from the tokens.

If you then ask: “Show me details for DESK400”, DESK400 in your message is also automatically replaced by ENTITY_000 before being sent to the AI service.