AI Assistant

The AI Assistant is an optional feature of the Allegro Packets Network Multimeter that runs a local Large Language Model (LLM) directly on the device. All processing happens on-device, using only CPU resources that are not required for core packet processing. No data leaves the system.

The Assistant supports a variety of use cases, including:

• General networking questions
  “What is a TCP zero window?”

• Natural-language summaries of statistics
  “What can you tell me about the collected statistics for this connection?”

• Contextual analysis of metrics
  “Are the TCP response times for this IP reasonable?”

• Ad-hoc data exploration
  “Which IPs sent the most traffic in the last minute, and how does that compare to the last hour?”

Requirements

To use the AI Assistant, ensure your system meets the following requirements:

• Memory: At least 32 GB RAM (total)
• Storage: At least 12 GB free disk space
• Connectivity: Internet access for the initial model download
• System load: Low to moderate load recommended
  (The Assistant uses leftover CPU resources. Under high packet processing load, responses may be delayed or paused.)

Activation and Configuration

The AI Assistant can be configured in the Global Settings (bottom section).
A processing restart is required for any changes to take effect.

Model

1. Click Select Model
2. Choose a storage location with at least 12 GB of available space
3. Click Download Model
   • The model will be downloaded from Allegro Packets servers (Download time varies on your network connection)
   • If the system cannot access the servers a manual download link will be shown and you must download the model on another device
4. Select the model and confirm

Reasoning Effort

Before generating a response, the Assistant internally performs a reasoning step to improve answer quality.

• Low (recommended): Fast responses, suitable for most use cases
• Higher settings: More detailed reasoning, but slower responses

Use higher settings only for complex analytical queries.

Fetch Statistics

The Assistant can automatically retrieve relevant statistics to answer your questions.

• This may increase response time, especially with large datasets
• Typically, statistics are only fetched when explicitly requested

Disable this option on smaller systems if consistent response times are required.

Usage

You can interact with the AI Assistant in several ways:

• Dedicated AI Assistant page:
  Ask free-form questions
• Contextual entry points (✨):
  Use these to Analyze displayed data or explain metrics.
  Currently only available in IP and Connections Table.
• Text selection:
  Highlight any text in the UI and ask a question about it

History

By default, conversations are not stored.

You can enable per-user chat history in the User Settings:

• Conversations will be saved per user
• Stored chats are accessible from the AI Assistant page
• It is possible to delete individual chats or clear the entire history