Building Compound AI Systems with Agent Tools and Function Calls

What you’ll learn

AI systems need a variety of skills to handle different business tasks and provide useful solutions for each industry. But having these abilities alone isn’t enough to really help your customers with their specific data. For the AI to give helpful answers, it needs access to details like a customer’s contract, their recent support emails, or the latest sales report.

This is where Compound AI systems come into play. In this demo, we’ll create a Compound AI system that combines several tools (called Unity Catalog Functions) and reasoning agents. These tools usually retrieve additional information or perform certain tasks, such as:

  • Checking the latest customer order status (for questions about an online order)
  • Fetching the current weather
  • Using an external machine learning model for recommendations
  • Searching a database to find relevant information or documents for better answers (RAG)
  • Running specialized queries or prompts with fine-tuned AI models
  • Running Python scripts
  • Performing mathematical calculations
Launch Product Tour

Recommended

<p>Build High-Quality RAG Apps with Mosaic AI Agent Framework and Agent Evaluation, Model Serving, and Vector Search</p>

Tutorial

Build High-Quality RAG Apps with Mosaic AI Agent Framework and Agent Evaluation, Model Serving, and Vector Search

<p>LLM-Tools-Functions: Build Your First Compound AI Systems</p>

Tutorial

LLM-Tools-Functions: Build Your First Compound AI Systems

<p>Quickly Build, Deploy, and Assess a RAG Application with the Mosaic AI Agent Framework and Agent Evaluation</p>

Product Tour

Quickly Build, Deploy, and Assess a RAG Application with the Mosaic AI Agent Framework and Agent Evaluation