AI for BI and Multi-Agent apps with OpenAI's Swarm
Generative AI can easily analyse data and create charts we develop a simple BI program. We also take a look at how to create multi-agent apps with Swarm.
I have a couple of topics for you in this newsletter: Business Information and graphics generated by AI, and multi-agent frameworks with OpenAI’s Swarm.
Multi-agent frameworks (Swarm)
I was intrigued by the large number of multi-agent orchestration frameworks that have appeared recently, including LangGraph from Langchain, AutoGen, CrewAI (that I looked at here), MotleyCrew, and several others that I have not had time to look at.
But when I saw OpenAI’s Swarm, I thought I should investigate. Swarm is not a sophisticated frame but a back-to-basics learning tool for creating multi-agent apps. The aim is not necessarily to build state-of-the-art apps but to learn the basics of how they are constructed using routines and handoffs. I’ve written a tutorial that explains how to build a multi-agent, tool-using AI app, here:
Hands on with OpenAI Swarm
AI for BI
LLMs have been able to write code and construct charts for some time now but I thought I’d revisit the idea with a specific application in mind. It’s a program that will construct a Business Information report by taking CSV data, building charts and writing a report about the data. It’s a simple idea but could be easily developed into a more sophisticated application. See my experiments here:
AI for BI: Building a Business Information Report with CrewAI and OpenAI - How to create a simple BI report directly from data
If you are interested in Data Visualisation and AI, Streamlit is a great way to get started. I’ve written a book to help you get started.
Streamlit from Scratch
Streamlit is a framework for developing data science and visualization applications. Streamlit from Scratch is a a book that will quickly get you up to speed developing those apps.