AIOS: LLM Agent Operating System! Create Software, Automate Tasks, and More!

[Music] we have been seeing more and more Frameworks being built to integrate AI into our operating systems today we're going to be taking a look at aios which is a project that is infusing AI into our own operating systems to automate as well as deploying AI agents onto our computer aios is a large language model agent operating system it's something that embeds different large Nish models like mixol you have quen as well as many other LMS into our operating system as a brain of the OS and it's something that enables the OS to have a soul you can say it's a system that is designed to optimize resource allocation facilitating context switch across different agents enabling concurrent execution of Agents providing tool services maintaining access control and providing Rich sets of toolkits for different large language model agent devs now I want you guys to take a look at this figure right here which is a figure that is motivating example of how an agent a travel agent in this case requires both large language model level and an OS level resource and functions to complete a task so in this example it's utilizing both of these two factors to function on completing a different task it involves organizing a trip based on the user's preference so we can see that the user is stating that I'm flying from San Francisco to New York for the for business next month please help me organize the trip so this this query is then sent to the travel agent that is operating on the OS so whenever the agent interacts with the large language model service for tasks such as retrieving preferences it makes decisions on tools as well as apis that generate reviews as well as responses it also interacts with the traditional OS services for tasks like accessing dis drives you also have the ability to execute software so for each step that you can see over here there's one step two steps three steps Etc and in these steps it involves a combination of a large language model reasoning as well as an OS level action so it focuses on the preference retrieval you have the flight and whel recommendation which is utilizing different things such as a tool API that is managed by the LM you have an LM storage managed by the large language model this storage which is operated by the traditional OS you have tool API software software text generation and all of these steps are working alongside side with the agents that are deployed on our OS sorry for being repetitive but this month we had insane Partnerships with big companies giving out subscriptions to AI tools completely for free these are tools that will streamline your business's growth and improve your efficiency just being a patreon this past month you were given access to six paid subscriptions completely for free not only do you access these subscriptions but you gain the ability for Consulting networking collaborating with the commun Community as well as with myself you get access to daily AI news resources giveaways and so much more if you're interested check out the patreon link in the description below to gain access to these benefits now if you are to read the research paper further it discusses the challenges that are posed by increasing complexity and the quantity of Agents such as the resource management of Agents scheduling as well as privacy concerns so what this paper has done to address these challenges is that they proposed the aios architecture which is depicted in this figure below and this is something that we're going to take a look at as we go further but in brief it's an architecture that includes a large language model specific kernel and it's designed to isolate large language model related tasks as well as resources from other OS functions and this kernel like is basically broken off into several different modules and it's something that is really interesting and it's something that you can actually Deploy on your computer today and it's something that we're going to be taking a look at as how you can get started so that you can deploy these different AI agents so that it can work on accomplishing various tools such as an agent scheduler that prioritizes and schedules agent requests to optimize large Lage model utilization and such forth I'm not going to be going over all of these cu the intro is already way too long but this is something that we're going to be taking a look at throughout today's video so I hope you enjoy it stay tuned and let's get straight to it if you would like to the book a one-on-one with me where you can access my Consulting Services where I can help you grow your business or basically give you a lot of different types of solutions with AI definitely take a look at the calendar Link in the description below hey what is up guys welcome back to another YouTube video at the world of AI in today's video we're going to be taking a look at this project which is aios large languish model agents in our operating system now we already mentioned how it basically functions but let's actually take a deeper dive as to how this architecture like operates it as I stated before it functions through kernels that consist of several different modules we talked about agent scheduler but context manager is something that supports snapshotting and restoring the immediate generation status and this is for the large language model it manages the context window as well you have memory managers which are providing short-term memory for agents that are in interacting with different logs you have store manager storage manager which is a persistent agent that interacts with logs in longterm storage tool manager it is a manager that agents call for external API tools and lastly you have the access manager which enforces privacy and access control policies between agents these basically are different modules that work together to enhance the management and the coordination of large langage model related activities Within in the operating system that you deploy aios within so this is how the overall like architecture would work based off these kernels now we're going to go into a deeper dive as to how it basically functions and this is where the kernel exposes in LM system through a call interface for agents to access these services so it basically functions through this whole architecture it's kind of complicated but if you are to get a deeper Dive by reading through the different layers which they specify you would get a better idea now before we go further into the video I want to specify for the people who would be interested in installing this it's kind of pretty easy what you would want to do is you would want to clone this repository so you would basically want to go into your command prompt and there's a couple prerequisites that you would need which I should specify before first you need to make sure that you have Pip get as well as a functional hugging face Hub token ID so once you have these things ready as well as the memory to host something like a gamma 2 uh two bilon parameter model you should be able to host this quite easily so what you're going to be doing is going into your command prompt you're going to be cloning this repo so what you would do is just clone this repo by copying it and clicking this green button which will give you the access to the link you can scroll down go into your repo and type in get clone press enter once you have cloned this you can then wait until it finishes cloning and then you can go into this a iOS folder by typing in CD aios once you're in this directory you can start installing the requirements and they basically specify that make sure you have python 3.9 or greater and you need to install the required packages with this command so you can simply just go in paste this command in and press enter this will take a couple of seconds to a couple minutes depending on your computer but once it finishes doing that in the meantime you can work on finding your hugging face ID which I'll showcase how you can do it so to get your hugging face access ID you can go on to hugging face if you do not have an account definitely create one once you have done that click on this icon over here which is your profile and go to settings once you are in your settings go to the access token key create a new token you want to name it a iOS and you can just keep it whatever you want generate the token copy this copen ah token and then we can go proceed forward with installation now what you want to do next is then set your hugging face ID within a iOS this is obviously after has finished installing the requirements so what you would want to do is copy this command now if you are on Windows you don't want to use the expert function you want to use set as your command and you would basically type in set huging face ID everything normal just replaced your uh token in this uh section over here which is your read token what once that is done you want to then do set hugging face and then your cach directory and once you have basically set that you can run the main.py to start and this is something that's really important you can choose between these two LMS at the moment you have gamma 2B as well as mistol 8x7 V so what you would want to do in this case you want to make sure that you replace the max GPU memory and the Evol device based off of your own requirements and you can definitely find that on your own operating system or your own control panel which will tell you the information on what's your max GPU I currently do not have any requirements to run these models so I'm not going to be testing it out I don't want to find my fry my computer but this is easy as that you can then have the interface which will pop up into your command prompt stating what you need to do next it'll basically specify what sort of task you want to do with the LM as the brain of your operating system and it can execute various tasks on your system and do you can basically accomplish various things with the different modules that we mentioned before and that's about it for getting started with aios so let's actually take a look at the architecture once again because we have a couple components that we have left to take a look at we have these aios layers it's basically split up into three layers you have the application layer you have the OS layer which is the kernel layer and then you have the hardware layer firstly with the application layer it's where the agent applications like the travel agent you have the coding agent math agent are being deployed and developed so what happens is that this system the aios system is providing the SDK at this layer which abstracts of the system calls are being used to simplify the whole development process for the agent so it's working alongside with the aios SDK it is then sending the related calls to the LM if it's not needed for LM based activities it is then going to the OS kernel which is then going to require the hardware functionalities and this is where we go to the kernel layer and it's basically consisting of the two main components which we mentioned before the OS kernel and the LM kernel the OS kernel handles the non-specific operations of an LM as whereas the LM kernel is dedicated for LM specific task and this segregation you can see is allowing for LM kernel to focus on the critical LM related activities and these are the modules that we mentioned they actually go a little bit more in depth on each and every module which gives you a better idea as to what you can do with these different agents when they're deployed on your OS and lastly we have the hardware layer this is encompassing a physical component of the system such as your CPU GPU memory as well as many of the other components over here now after reading couple of sections of this report I was just wondering what are some practical use cases for aios like why would you actually involve having the trouble to integrate AI into your operating system well there's obviously going to be a lot of different use cases for different preferences but one thing that would really benefit for me is that having it so that it would be used for coding and this is where I would deploy different AI powered large linguage models by infusing it into the operating system based off the framework that they have provided and this would involve deploying a personalized virtual coding assistant and this would help me for like basic coding needs filling out different prototypes as well as helping me in various cases where I would need debugging now this coding agent would be tailored for specific programming languages and it would also help me with realtime support and that was just one thing that I I've really thought about cuz it could help me with auto completion code Snippets it could help me with airor detection as well as syntax highlighting and this was one thing that I was just thinking that it could assist me in and I would just wonder what other people would do with this and I really always whenever I make videos on this I know these tools are really helpful but in Practical use cases I don't really try each and every one of them out but I would really be intrigued to see what my viewers are building and doing with these different tool sets so if you're interested you can even pitch me an email showcase what you've built with it and that I can like possibly even make a website in the future where I compile all the tool sets that I showcase and all the US cases that people have developed and this is like the community aspect that I really want to develop with this Channel I want to have an interactive AI community that would be basically pitching out different ideas and different tools I already do that with the Discord but I would be interested in having a more open-source like Community where everyone can contribute to that so if you're interested in that definitely send me an email on what sort of things that you would build with all the tools that I cover in my like videos and I'd be interested to see what you guys have been building and obviously if it's super interesting I would definitely make a video on it but I know I'm rambling too much now that's basically it for today's video I hope you enjoyed it guys this is definitely an a great tool that I personally see some sort of use case out of for deploying on your local computer it's basically giving a brain to your OS and this is something that we have mentioned previously in this video so I hope you enjoyed this video guys if you haven't definitely check out the patreon page this is a great great way for you to stay up to date with the latest AI news as well as accessing different subscriptions we have a growing community and it's just growing like tremendously over the past few months and I truly recommend that you would love it and you if you are to join it you would definitely benefit from make sure you check out the Twitter page if you haven't this is a great way for you to stay up to date with the latest AI news and lastly make sure you guys subscribe turn on the notification Bell like this video check out our previous videos there's a lot of content that you would would definitely benefit from so with that thought guys have an amazing day spread positivity and I'll see you guys fairly shortly peace out fell

Channel: WorldofAI

Views: 11,201

Keywords: AIOS, LLM Agent Operating System, Artificial General Intelligence, AIOS Architecture, open ai, artificial intelligence, copilot, os-copilot, semantic memory, ai agents, ai agent, you personal ai agents, rabbit r1, llm, ai, ai agents tutorial, ai agents explained, multion ai, ai personal agents, personal ai assistant, ai assistant, chatgpt, ai technology, ai for productivity, ai application, digital assistant, agent, autogpt, microsoft ai, microsoft copilot, generative ai

Id: 7WVP0yWVrLI

Length: 15min 7sec (907 seconds)

Published: Fri Mar 29 2024