DominoIQ - An Overview

1.1 What is an LLM (Large Language Model)?

• An advanced AI system trained on massive amounts of text data

• Answers questions and summarizes texts

• Generates creative content

• Provides support for tasks through natural language processing

• Its performance is largely determined by the choice of the right LLM

1.2 EU AI Act - Important for Companies

• EU regulatory framework for controlling AI technologies

• In force since August 1, 2024

• Mandatory AI training for employees by February 2, 2025

• Risk reduction through better training measures

• Competitive advantages through competent use of AI

3.1 Domino Directory Profile Setup

• Central configuration of all DominoIQ servers within a Domino domain

• Setup directly in the Domino Directory Profile (names.nsf)

• Definition of the responsible administration server

• Registration of all servers running DominoIQ

3.2 Database Setup

• dominoiq.nsf is automatically created by Domino

• Serves as the configuration database for one or more servers

• Must be present on every DominoIQ server

• Contains all required configurations, commands, system prompts, and models

3.2.1 Model Configuration

• Automatic download and management of LLM models in GGUF format

• Done via an "LLM Model" document in dominoiq.nsf

3.2.2 System Prompts

• Creation of a "System Prompt" document

• Instructs the DominoIQ AI inference engine on how to formulate queries

• Helps generate precise answers

3.2.3 Commands

• Assignment of LLM system prompts to DominoIQ servers

• Creation of an "LLM Command" document

3.2.4 Configuration

• Creation of a "DominoIQ Configuration" document

• Links a Domino server with an LLM model

3.2.5 Remote Configuration

• Creation of a "DominoIQ Configuration" document in remote mode

• Allows connecting a Domino server to an externally hosted LLM endpoint

3.3 Model Download

• Download your preferred LLM (huggingface.co is the largest hub for LLMs)

• DominoIQ (LLAMA server) requires models in GGUF format

• The user interface provides options for selecting LLMs available as GGUF files

• Recommended for initial tests: Llama-3.2-3B-Instruct-GGUF

3.3.1 Recommended Models for Different Use Cases

3.4 Additional Configuration Steps

3.4.1 Model Configuration

• Name of the model

• Short description of the model

• GGUF model file name on the local drive

• Storage location of the GGUF files: "/DOMINODATA/llm_models" directory

3.4.2 Server Configuration

• Creation of a configuration document for each Domino iQ server

• Definition of the LLM model to be used by the server

• Configuration of the port for the localhost connection

3.4.3 Remote Server Configuration (optional)

• URL of the external LLM server

• API key for connecting to the LLM server

• Manual generation of a secure key value (bearer token for the REST API)

3.4.4 TLS Configuration (optional)

• DominoIQ offers full TLS support

• Uses CertMgr TLS credentials

• CertMgr now supports "localhost"

• Localhost certificates can be created with MicroCA