Name: Bee Agent Framework
Uploaded: 2025-01-20T13:22:40.559Z
Duration: 2 min 33 s
Description: An open-source framework for building, deploying, and serving scalable agent-based workflows with various AI models.

bee-agent-framework

An open-source framework for building, deploying, and serving scalable agent-based workflows with various AI models.

Bee Agent Framework

Hello everyone. Welcome to the channel. In this video I'm going to show you how you can locally install this newly released very active project called as Bee agent framework. Bee agent framework is an open-source framework for building, deploying, and serving powerful agentic workflows at scale. The Bee agent framework makes it really easy to build scalable agent-based workflows with your model of choice, and it works natively quite well with AMA. 

The framework is designed to perform robustly with IBM's Granite models and Lama 3.2 or 1 models. This project is quite active in the sense that they are also optimizing it to work with other popular large language models. I believe that this model has been released by the IBM developers, and this is primarily geared towards the open-source Community, which is quite heartening. 

I'm very, very glad to see that how well IBM is progressing in order to service the open-source Community when it comes to AI. Their Granite models I already have covered them in detail. You can search them on the channel. They're also of quite high quality. The basic aim of this project is to empower the developers to adopt the latest.

Open source and proprietary models with minimal changes to their current agent employ implementation. You can use powerful Bee agents which are refined for Lama 3.1 or you can even build your own. You can use the built-in tools or you can create your own tools in Python JavaScript. It also gives you a code interpreter which is a Sandbox container. 

You can adopt a lot of strategies for keeping the memory because look what is happening these days is that these agents are autonomous pieces of software which are asynchronous which use the power of AI. They take input as text prompt or natural language and then they decide what needs to be done. You can add memory to them so that they remember what they did last time and then you can orchestrate them in various ways. 

This project aims to support all of it. It already is trying to do a lot of these things but, uh, I believe that it still needs a bit more time to get more mature and I'm more than sure it will because the start looks quite promising. 

Okay, enough talk. Let me take you to my terminal and we will get cracking on this one. Before I do that, let me give a huge thanks to Mast compute who are sponsoring.

The VM and GPU for this video and the GPU is needed because we will be using AMA based local models. So that is why I will be installing them locally. If you're looking to rent a GPU on affordable prices, I will drop the link to their website in video's description. Plus I'm also going to give you a coupon code of 50% discount on range of GPU, so do check them out. 

Okay, so that said and done, let me show you my terminal where I'm running this 12 22.041 GPU card Nvidia RTX a6000 with 48 GB of vram courtesy MK compute. So first up, allow me to create a virtual environment and we are going to do everything in that environment. Also, another prerequisite you would need is npm and node install, so make sure you have it. If you don't know how to do it, just search my channel. I have done a very easy to understand video and you can see that I have this npm version installed and I have this node installed, so that is good. 

Also, I already have AMA running here. AMA is installed. If you don't know what is how to install it, please search my channel. I have done like more than 500 videos on it or maybe more in the last one year or so. So in order to install all

You need to do is to click on this download and then download it according to your operating system. Now the model which I'm going to run with it, which is a default model, is Lama 3.1. This one in 8 billion, so it's already selected. So I'm just going to copy it from here and then I'm going to paste it here. It is going to download this model on my local system. 

And once it is downloaded, it is going to verify the check sum and then model will be served on our local system easily. This is how easy it is to run this AMA based stuff here. So model is already downloaded and next step is that it is going to do the check some verification, which shouldn't take too long and that should be ready any second now. 

And my local model is loaded. Let me ex it from here and if I do Lama list now, you see that my Lama 3.1 is running here. You can use any other model if you like, but go with Lamar grite at the moment they are more optimized with it. Next, let's install this B agent framework with npm. All you need to do is to run this command and it is going to download and install all the required packages for this P agent framework. So let's wait for it and B.

Agent framework is installed. Took around 1 minute and now let me show you an example of it. Okay, so in order to run it, let me show you the code which you would need to run to create your agent with this framework. I'm just opening it in my VSS code on my local system. Just create a new file and name it anything but with a MGS extension that it's an es module in node. Then import this B agent framework AMA and then Dr goet because agents use tools. 

So we are just using this web search and then we can also use this open uh materal tool for uh for weather and all that stuff. We are instantiating our llm and you can see that by default it uses Lama 3.1 running locally. But if you are using it in production environment, go with 70 billion model. 

Then we are creating this agent. We are just initializing its llm which we have defined our memory and the tools which this agent can use are web search and for weather. You can there are various tool which you can find there uh on their GitHub rep and I will drop the link to it in video's description. Then you can just simply go with uh run the response in the async mode. We are asking it what is the

Current weather in La Las Vegas and then it is going in checking the weather and locking it. So let me run it. I'm just going to say node and then Bee agent.

MGS let's wait for it to run there you go. So agent is thinking that okay I'll just go up and show you what it did. 

So it says agent thought is the user want to know the current weather in Vegas. I can use this function to get the current weather. So agent has automatically selected which tool it can use. From there it has given it an input Las Vegas country USA start this this this and then in foreen height and the tool caption is this. 

And then it has got all the information from that tool. And then agent thought is a current weather. So agent has automatically processed all that information from above. And then it is telling us the current weather in Vegas is mostly sunny with a high and low and final answer is this. 

And then agent has return return the answer to the user. How good is that really really awesome and neat stuff. In the next example you can search anything with this agent. For example, if I want to ask what's the nearest tire shop to Sydney CBD and then if I go here and maybe run it let's see what it does. 

Agent thought user F wants to find out. And if I just go up sorry quickly it says the user wants to find a tire shop near CBD.

Can search online for that Duck Duck Go and then it is just telling me about that how good is that all the title and stuff. It is gathering the information and of course it can you can uh you format your prom that you want to know the cheapest or the something like that. 

And then it is telling I know the final answer and all that stuff but you know so you can see that randomly you can and just ask anything you like. If it is related to your own application you just need to modify that tools and that's about it. You can define a function replace it with this tool call from here that's all you need to do and you can import it as you like. 

As I said it is a very early project but still very promising. As you can see it's very easy to create agents it is very easy to run agents. I'm very impressed by it so heads off to IBM for creating this. 

Let me know what do you think about this project. Before I let you go let me also introduce you to the sponsors of the video who are agent ql. Agent ql is a cury language that turns any web page into a data source. With its python STK and live debugging tool you can scrape and interact with web content.

Agent ql works on any page. It is resilient. It is reusable and it structures output according to the shape of your cury. I will drop the link to their website in videos description so do check them out. 

Note, that there may be slight differences in text and video. Watch video for the original content.

Bee Agent Framework Video