As an enthusiastic open source developer, I was thrilled to get the opportunity to become a speaker at Open Source Summit Europe 2024, which took place in Austria (Vienna - my hometown) on September 16-18, 2024.

What is OSSEU?

Open Source Summit is the premier event for open source developers, technologists, and community leaders to collaborate, share information, solve problems and gain knowledge, furthering open source innovation and ensuring a sustainable open source ecosystem. The initiator of Open Source Summit is the Linux Foundation. It is the gathering place for open source code and community contributors.

Motivation

For the last year, I have been immersed in the world of generative AI and mainly large language models. But in the past few months, I have been fully dedicated to the open-source project Bee Agent Framework, which is the core part of the Bee Stack.

Bee Agent Framework

An open-source framework for building, deploying, and serving powerful agentic applications at scale. The framework makes it easy to build scalable agent-based workflows with your model of choice.

Let's try it on your own with the freely available Llama 3.1 8B model, which is accessible through ollama.

# Download the model
ollama pull llama3.1:8b

# Init a project
mkdir bee_agent_framework_demo && cd $_
npm init -y && npm pkg set type="module"
npm install bee-agent-framework ollama

agent.js

import { BeeAgent } from 'bee-agent-framework/agents/bee/agent'
import { OllamaChatLLM } from 'bee-agent-framework/adapters/ollama/chat'
import { UnconstrainedMemory } from 'bee-agent-framework/memory/unconstrainedMemory'
import { OpenMeteoTool } from 'bee-agent-framework/tools/weather/openMeteo'

const agent = new BeeAgent({
  llm: new OllamaChatLLM(),
  memory: new UnconstrainedMemory(),
  tools: [new OpenMeteoTool()],
})

// Let's ask the agent two subsequent questions
for (const prompt of ["What is the name of the capital city of Austria?", 'What is the current weather there?']) {
    console.log(`User 👨‍💻: ${prompt}`)
    const response = await agent.run({ prompt })
    console.log(`Agent 🤖 : ${response.result.text}`)
}

and then run it

node agent.js

To see more examples, visit the Bee Agent Framework on GitHub.

The talk

Based on the gained experiences, we have decided to go public, and the best way of getting traction is to talk about it publicly and the OSSEU was where we started.

The title of my talk was Towards Useful, Reliable, and Scalable AI Agents and had the following abstract.

AI agents have the potential to unlock the next wave of enterprise productivity, but implementing them in production is a complex task. In this session, we'll delve into the key challenges of building agents that are not only useful but also safe, reliable, and scalable in real-world applications. We will discuss and demonstrate with code the essential components of building AI agents and introduce the Bee stack, an open-source initiative by IBM Research to empower developers to build robust agentic applications.

Agenda

1. From standalone LLMs to LLM Agents (1:43 - 10:57)
1. Overview of key challenges with LLM agents (10:58 - 18:18)
1. Introducing the Bee Stack (18:19 - 21:45)
1. Working with the Bee Agent Framework (21:46 - 38:38)

Recording

Slides

Unfortunately, the code snippet animations are not working in the embedded slides but are clearly visible in the recording.

Conclusion

Showing in conferences like this one is always great. You gain some knowledge, meet new people, and exchange your point of view, and Open Source Summit Europe was no exception.

What's next?

In the upcoming weeks, I would like to provide a deeper dive into the framework itself and the other parts of our Bee Stack. But for now, do me a favor and try the Bee Agent Framework. I am looking forward to seeing your feedback!

LLM Agents: My Talk at Open Source Summit Europe 2024

Tags