Tutorial: Multi-Agent Concierge: How to Build An Agentic System (Stock/Finance Agent)

??Try it on Google Colab

Interactive chatbots have become a familiar solution in customer service, using agents to provide memory, introspection, tool use, and other features necessary for a competent bot. However, as the complexity of tasks increases, traditional single-agent systems can become unwieldy and difficult to manage.

This is where multi-agent systems come in, particularly in environments where tasks are interconnected and dependent on each other. For example, a bank implementing a system that can:

• Look up the price of a specific stock
• Authenticate a user
• Check an account balance (which requires authentication)
• Transfer money between accounts (which requires both authentication and balance checking)

Each of these top-level tasks has sub-tasks. For instance:

• The stock price lookup might need to look up the stock symbol first.
• The user authentication would need to gather a username and password.
• The account balance inquiry would need to know which of the user's accounts to check.

Designing a single primary prompt for all these tasks and sub-tasks would be extremely complex. Instead, a more modular approach is needed—a multi-agent system where each agent is responsible for a specific top-level task, plus a "concierge" agent to direct users to the correct agent.

Example of a Multi-Agent System

In this tutorial, you'll learn how to build a system of agents to complete tasks in a financial environment. The system includes four basic "task" agents and three "meta" agents:

Task Agents:

1. Stock Lookup Agent: Manages stock price lookup.
2. Authentication Agent: Handles user login.
3. Account Balance Agent: Retrieves account balances.
4. Money Transfer Agent: Facilitates money transfers between accounts.

Meta Agents:

1. Concierge Agent: Guides the user and directs them to the appropriate task agent.
2. Orchestration Agent: Decides which agent should handle the next step based on the user's current state and request.
3. Continuation Agent: Chains tasks together by determining what the original task was and ensuring it is completed, even if multiple agents are needed.

??Try it on Google Colab

?? GitHub GIST

Tutorial Overview

Core Components of the System

1. Install and Configure Requirements

• Install necessary libraries.
• Set up your OpenAI API key.
• Check that everything is properly installed.

Multi-Agent Configuration

• Define and implement each of the core agents (Stock Lookup, Authentication, Account Balance, Money Transfer).
• Set up the meta agents (Concierge, Orchestration, Continuation) to manage task flow.

Running the System

• Initialize the system and run the main loop.
• Manage user input and direct tasks to the appropriate agent.

Advanced Usage and Customization

• Adding new agents to extend the system.
• Customizing agent behavior and integrating them into the orchestration logic.
• Creating interactive configuration UIs within the Google Colab environment.