magnetic one a generalist multi-AI agent framework created by Microsoft now you can use it with Ollama magnetic one agents includes web surfer agent file surfer agent coder agent computer terminal agent and they all have access to different tools across various

benchmarks using this generalist agent the accuracy is comparatively higher when compared to using GPT-4 on its own the orchestrator agent is able to get the task create sub tasks assign those tasks to other agents and get the work done even though the task is complex it

also has the capability to browse the web click on icons as you can see here it's clicking icons to complete the task so I'm going to take you through step by step how you can install this configure magnetic one agents with Ollama and then

run it but before that I regularly create videos in regards to Artificial Intelligence on my YouTube channel so do subscribe and click the bell icon to stay tuned make sure you click the like button so this video can be helpful for many others like you step one install

first install Docker on your computer then install the UV package manager using this command just copy it and then paste it in your terminal and then click enter to install it next go to llama.com and download Ollama and finally go to Bing web search API and create an API key I

have already added Ollama to magnetic one and asked to get merged to AutoGen so hopefully it'll get approved soon but before that I'm going to clone from my own repo so this might change in the future so get clone this URL and then

click enter next navigate to autogen / python folder there we are going to use pip to install all the required packages once off it gets installed just source .env / bin / activate then navigate to packages autogen magentic 1 here pip install hyphy dot and then click enter

now this will install all the required packages I have clearly explained step by step in my blog post includes installing all the packages and the steps involved step two configure export your Bing API key like this and then click enter now you need to type O llama

pull llama 3.2 hyphen Vision to download the vision language model now it got installed now we need to set the chat completion provider to olama by typing like this and then click enter similarly we need to add our chat completion keyword arguments in Json format

providing the model name and the base URL this is the olama base URL after this click enter now configuration step is complete step three running the application just type python example this whole command and then click enter now it's asking me to enter the task so

I'm going to say get the stock price of Apple using Y finance and give me the percentage change now it's processing the request you can also monitor this in the my logs folder there you can see the screenshots being taken here's the next screenshot the web Surfer agent tried to

get the information but it was not able to fetch those information so now it goes to the coder agent and the coder agent is writing the code seems like the orchestrator didn't run the code but it's hallucinating with these values this could be because l 3.2 is not as

effective in regards to function calling, and we use the 11 billion parameter model. Maybe if I try a 90 billion parameter model, this should have worked. Finally, we managed to make it work with Llama. I'm going to create more videos similar to this, so stay tuned. I hope you

like this video. Do like, share, and subscribe, and thanks for watching.

Run Microsoft Magentic AI Agents with Ollama in 5 Minutes or LESS!

Mervin Praison

Magnetic One은 Microsoft에서 만든 다중 AI 에이전트 프레임워크입니다.

여기에는 웹 서퍼, 파일 서퍼, 코더 및 컴퓨터 터미널 에이전트와 같은 에이전트가 포함됩니다.

GPT-4만 사용하는 것에 비해 더 높은 정확도를 보여줍니다.

오케스트레이터 상담원은 하위 작업을 관리하고 다른 상담원에게 할당합니다.

상담원은 웹을 탐색하고 아이콘과 같은 요소와 상호 작용할 수 있습니다.

Docker, UV 패키지 관리자, Ollama를 설치하고 Bing 웹 검색 API 키를 받습니다.

리포지토리를 복제하고 지정된 폴더로 이동합니다.

pip를 사용하여 필요한 패키지를 설치합니다.

Bing API 키를 내보내고 Llama 3.2-Vision 모델을 가져와서 구성합니다.

채팅 완료 공급자를 Ollama로 설정하고 키워드 인수를 추가합니다.

애플리케이션 실행 및 문제 해결

애플리케이션을 실행하고 작업(예: Apple의 주가 가져오기)을 제공합니다.

프로세스 및 로그를 모니터링하고 상담원 상호 작용을 관찰하세요.

문제 해결: Llama 3.2 및 함수 호출과 관련된 잠재적 문제.

해결 방법: 더 큰 매개변수 모델을 사용하면 결과가 개선될 수 있습니다.