Skip to main content

Overview

Witty is a personal assistant designed to enhance user productivity by leveraging Large Language Models (LLMs) and Generative AI (GenAI). It operates as a multi-agent system, dynamically selecting the most suitable AI agent based on the user's request.

Concepts​

Traditionally, an agent refers to a system that can independently act within an environment to achieve a specific goal. The emergence of large language models (LLMs) and business logic models (BLMs) has expanded this definition, introducing software agents that leverage foundational models to make decisions within a workflow. For more information about AI agents read Agentic AI: The Next Evolution in AI and Data-Driven Decision Making

Witty works as a supervisor that independently chooses the best AI agents it needs to achieve those goals.

Witty schema

By default, Witty includes a specialized agent with in-depth knowledge of Witboost documentation, enabling users to quickly access accurate and context-aware information.

Other specialized agents can be added at any time by platform team.

Components​

Witty is made of three main components:

  • Agents catalog: collection of AI agent configurations. How to add a new agent is described here;
  • LLM configuration: user-defined LLM model used from supervisor. Also, it can be accessed by other AI agents via API call. How to configure a LLM model is described here;
  • Embedder configuration: user-defined embedder model. How to configure an embedder model is described here.