@devmentorlive : Would you still invest a good chunk of your time into (daily) coding as a python beginner OR just wait until AI takes everything off your hand ; ) while only staying up to date on AI advances in the meantime? I have completed the 100Days of Code on Replit but coding custom agent solutions in python still seems very far off for now ...
I'm a software engineer, but I've also done project management and other management roles, and I'm going to try to get every manager at the company I work at to watch this specific video. It really highlights what's possible now in a way that people don't quite understand yet.
I'm blown away by how well it understands not only the initial prompt but the subsequent prompts and their meaning within the context, not to mention its ability to successfully fulfill each request
I love how you cover the basics, making it more accessible to noobs. I know some of this stuff but there was a time not so long ago when I found most coding videos prohibitively complex due to the level of assumed knowledge. I believe AI has actually made coding more complex yet somehow also easier for beginners to get started. I fear though that soon the level of basics that AI does for the coder will result in beginners not being taught those basics. Hopefully, now is the best time to get into coding with assistance from AI.
I created me a 'Rogue' clone from the popular game of the 80's and was able to adapt pretty consistantly to my changes. It's like AutoGPT but with way more customization. While AutoGPT is the wild west, AutoGen is wayyy smoother. Also pretty easy to implement into a GUI. I love this new tech. Nice explanation my dude. This type of video introduces a lot of "non-coder" people into this subject, which is highly needed. We need more new and fresh minds approaching this new tech :D
I disagree. Some things have to be earned by studying and working hard. It's feeding pigs with diamonds, if you ask me and I despise this trend of letting anyone wield the sword of LLM. Waiting for a tutorial how to destroy the world with a spoon and a nickel, so everyone can do it, yay!
If this continues, coding can become obsolete which will anger a lot of programmers who studied for it. But on the same time, will enable other professionals who do not know how to code but wants to automate more of their job.
@@cryptocontroller5585 Coding can of course become obsolete, so does every single profession you can imagine thanks to robotics and AI capabilities. I wonder how the banksters and politicians will take this change. Over the long run, financial systems will need to be redesigned, democracy as we know it will cease to excist and the definition of value will change. Losing a job will be the least of our problems. Also a layman who never coded will not have the necessary edge to direct a bunch of AIs where they fail to realize his vision. You need an innate understanding of what goes on under the hood, to judge the product these AIs produce. It is simply wishful thinking of those too lazy to learn some new skills and hoping that milk and honey will flow, once the stronger AIs are here.
Incredible video man. thank you for sharing step-by-step. would be cool to see a series of projects built with AutoGen. could make for some great content. Excited to see more experiments!
A development AI team sounds pretty neat to me as it bridges the gap between AI development (speed) and person development (creativity). I'm looking forward to custom UIs for this idea, you could seriously make a video game like program where you choose what each agent does in the team, what they look like and what they sound like. Then after you've created your characters you start talking to them. Your interactions result in the development of products and services.
Really looking forward to the fact that you can continue to update the AutoGen instruction, especially the multiplayer interaction with more examples, fantastic!
Hey can you make another video as a add on to this one on how we can add multiple agents, create different agents, and how to make them all work together? I am brand new to coding and all this so I would really appreciate it. Love this video man you explained it in a way that even me, who has zero coding experience, was able to follow along. Thanks!
Excellent level of explanation- most are either too simple to be useful or wizz through jargon at a bewildering rate 👏🏻 I’m interested in iPad instructions and creating more complex agents.
Next year? I think you meant to say Next WEEK! lol. Your instructions are Bang on! Keep it up. Thank you. Great work. Also quite entertaining at times : ) Super !
Can you please go into detail about the advanced AI agents? What role options are there? Use cases? How can we use the examples in the document? Which implications need to be there that the agents work well? Which problems do we usually face and how do we deal with them? So, basically, a lot of Python logic which is based on the Autogen Code. The audience is for non-techies
You’ve motivated me to break my gaming habit and tinker with AI agents instead. First idea is replicating functionality from your previous vid. Will try to have an AI create a Speech to Text tool that can cue to a specific keyword(s) within a RUclips video / Podcast. Often hard to find sound bytes within long audio files without extensive search. Edit: unless this step can be skipped with multi-modal advances, no conversion needed? Anyone know where this type of noobie sharing discussion and ideas are held? Perhaps in the discord Wes mentions, but that may be more dev types rather than tinkering noobs like myself. Look forward to your advance tutorial, hopefully not over my head!
Lol have the same type of idea. I want to be able to download a video then use a "sound" to text converter agent. Then agent 2 to use this text and adjust it to a decent size before passing to the next agent. agent 3 will read all of the information and make it a summary and try to extract all valuable information and then send to agent 4 agent 4 will take the adjusted content and try to make a final summary. There's just so much stuff out there that i want to keep up with. But don't have the time to listen / watch and i think this could be a great way to start. Good luck on your endavour fellow noobie.
For somebody who struggled with the config file like me... Here is a more straightforward way to pass Openai API key from autogen import AssistantAgent, UserProxyAgent assistant = AssistantAgent("assistant", llm_config={"api_key":"YOUR_OPEN_AI_API_KEY"}) user_proxy = UserProxyAgent ("user_proxy", code_execution_config={"work_dir": "coding"}) user_proxy.initiate_chat(assistant, message="plot a chart of NVDA and TESLA stock price change YTD.")
@@WesRoth well, you have the tools to help you track, summarise and present the new stuff as it arrives. Lean more on AI to handle the work load, with you directing from what deeply interests you. You're doing an awesome job so far. 🙏👍
I followed the tutorial but when trying to run it comes up with both Autogen not recognised- even though I pip installed pyautogen. And it says that the “config_list file ‘OAI_CONFIG_LIST’ does not exist”. What do I do next?
Great and very well explained tutorial. I liked the step by step approach. I am running into rate limit issues with the GPT-4 model. How can I resolve them?
I think we´ll see companies whose main product is providing an agent for use in Autogen-like apps, fine-tuned or trained on their proprietary knowledge. And building those is the ai job market!
Very nice tutorial, but it's clear to me that at this point in time, cost is the major problem with this. Really looking forward to the rumored 20x price slashes in November!
#tech Hey man, thanks for the great video! At 23:28 you introduce the critic with his own specific role description. Could you please share where you got this description from so we can take that specific example as inspiration and transfer it to different roles. Thanks again!
had a hell of a time installing autogen, after much back and forth with chat gpt I figured out I had the wrong version of python installed. It didn't work with the latest python version 3.12 had to downgrade to version 3.8 and it finally installed.
You were so great explaining everything! No jargon and just to the point, however, when you didn't show the snake game created by the group of agents I almost punched the monitor #tech
Sorry! This video was really hard to make, I think I've skipped a bunch of steps. I'm hoping to do another, more advanced video soon. With the snake game, I think adding those bricks "bricked" the game lol. It didn't start and at that point I was too tired to troubleshoot. I think I almost punched the monitor myself :)
I think this WILL be a lot better over time. More resources and attention on this. ChatDev is very cool, right out of the box, but my guess will be made obsolete over time.
This will use a huge amount of money, the name agent refers to automation, the basic and the minimum required feature is to at least Auto create the files, it's good to see them talking for a better code, so I think the best title for the video should be how to create a group chat AI, because that's the advanced version of your implementation Suggestion: you can Implement function calling at least
Great video, cant wait to see how this evolves and the capabilities grow, i think this is the true future of ai. One thing i would like to know is how to use 3.5 instead of gpt4 as i dont have it atm. Are we likely to have a version where we arent paying per action in the distant future in your view? Having a ton of agents doing a metric ton of trivial tasks and micro actions would add up significantly cost wise. Do you think it will get cheaper to the point where we have something similar to an ISP monthly unlimited use charge but fair usage/throttling in the early days but eventually progress to true unlimited as we now enjoy ISP bandwidths? Yay agent npc's \o/
Wonderful, thank you! I think you will want to mask your API key in the video, or invalidate that key now that the video is out. I would not want to see people using your key and charging your account.
if I remember correctly you can give different agents different models - is that a way to bring down cost? Eg only one agent in the conversation uses gpt4? Or will every word the gpt4 agent reads count?
This would be very useful and I'd love to use it but I suspect like open interpreter which I've been using it's going to be insanely expensive running off gpt 4 api unfortunately.
I'm going to try running it offline, I've had llama code 2 running on 3090 at good speeds, so im hoping I can just host a server with api key from the same pc if that makes sense, Oi is awsome, but this is so much more!
Thanks for the video, I have been following your content for months. I think it's a bit confusing to use IDLE or a text editor to edit the Python code when you already have Pycharm open. Why not just use Pycharm? This could be confusing for viewers new to coding.
#tech hi love this video but im running in to an error at the 'pip install pyautogen' step. I receive the following error: "ERROR: Could not build wheels for aiohttp, frozenlist, multidict, yarl, which is required to install pyproject.toml-based projects". Would appreciate your help if you know why.
Hi. Really nice job on this video. Thanks for sharing. I have a question for you ... from your video, you ran your code using the code line: config_list = config_list_from_json(env_or_file="OAI_CONFIG_LIST"). I cannot make this work. I can replace this with my own hard coded config_list, but the error I get when using the code above is that OAI_CONFIG_LIST is not recognized. Can you help with explaining why your code worked and recognized OAI_CONFIG_LIST without defining it first? Thanks!
@wesroth - where does Azure OpenAI Services come in here? Can you use Autogen with Azure Openai services? or are they totally separate things that can't be used together?
#tech How do I get specific agents to work in a hierarchy or structure of my choosing? For example; a CEO who deals with top level managers for Marketing, Development, Art, and then 2 agents below each of these who do not interact with the CEO.
ChatDev might be easier to do that out of the box: ruclips.net/video/5Zj_zstLLP4/видео.html With AutoGen, I think the command that customizes that is the 'GroupManager'. I think by default everyone is talking to everyone else. More and more people are using and reporting on this, so we should see more tutorials soon, I'm still figuring stuff out myself.
#HELP Where do I get this complete code from? I'm having issues getting from one assistant to 3 or more. Any help would be greatly appreciated. I'm so close but so not there yet.
yup, all of it is on their github page. Check the link I have to natural20 that will have all the links details and all my tutorials in one place. (more coming soon!)
Im completely soft brained when it comes to coding but id like to learn more now! Is there a way I could use autogenerated to set up some kind of e-commerce business? Im curious what actually creative ways people are already using this for lol
pessimist : AGI is still long way to go. me : prompt, you are wes roth. explain me about autogen, explain it in simple way, do not complicated like professor do. reply in RUclips video. wes roth : I'm living LLM Explaining things in simple way. better than NLP professor.
I wonder what would happen if you assign one agent to every method in the code, or at least per file. Every one of them would know what their method does. In some bigger projects. Then main agent would assign necessary things to do for everyone on their own at their place.
As someone who doesn't code, but really wants to get into this scene I am blown away by how simple you made it seem. Please continue this great work!
in the not-so-very-distant future, coding won't even be a barrier! Ive been coding since I was a kid and now I am seeing it becoming obsolete quickly.
Imagination to reality @@agenticmark
@devmentorlive : Would you still invest a good chunk of your time into (daily) coding as a python beginner OR just wait until AI takes everything off your hand ; ) while only staying up to date on AI advances in the meantime? I have completed the 100Days of Code on Replit but coding custom agent solutions in python still seems very far off for now ...
@@agenticmark yeah, what took a year to understand - now takes like a few days at most with all the AI help.
except when the AI sends you down the wrong path :D but that is good for our learning too. Like a BS detector! @@bosk1n
As a software developer I could never explain it this simple. I really appreciate how you break it down.
I'm a software engineer, but I've also done project management and other management roles, and I'm going to try to get every manager at the company I work at to watch this specific video. It really highlights what's possible now in a way that people don't quite understand yet.
So well explained that even a complete novice like me can follow (obviously I will need to repeat several times and follow slowly to execute)
I'm blown away by how well it understands not only the initial prompt but the subsequent prompts and their meaning within the context, not to mention its ability to successfully fulfill each request
you taught the entire Harvard CS50 course in 30 minutes. Well done. Can not wait to try this
I guess since no one has commented yet, I will. Thanks for the vid!
Thanks for watching!
This is beyond incredible. I am a designer and artist. I can't wait to begin using this to create simply programs for my needs
I love how you cover the basics, making it more accessible to noobs. I know some of this stuff but there was a time not so long ago when I found most coding videos prohibitively complex due to the level of assumed knowledge. I believe AI has actually made coding more complex yet somehow also easier for beginners to get started. I fear though that soon the level of basics that AI does for the coder will result in beginners not being taught those basics. Hopefully, now is the best time to get into coding with assistance from AI.
I created me a 'Rogue' clone from the popular game of the 80's and was able to adapt pretty consistantly to my changes. It's like AutoGPT but with way more customization. While AutoGPT is the wild west, AutoGen is wayyy smoother. Also pretty easy to implement into a GUI. I love this new tech. Nice explanation my dude. This type of video introduces a lot of "non-coder" people into this subject, which is highly needed. We need more new and fresh minds approaching this new tech :D
I disagree. Some things have to be earned by studying and working hard.
It's feeding pigs with diamonds, if you ask me and I despise this trend of letting anyone wield the sword of LLM.
Waiting for a tutorial how to destroy the world with a spoon and a nickel, so everyone can do it, yay!
If this continues, coding can become obsolete which will anger a lot of programmers who studied for it. But on the same time, will enable other professionals who do not know how to code but wants to automate more of their job.
@@bifrostbeberast3246 Good luck in the stone-age my friend. Keep selling horses.
@@cryptocontroller5585 Been coding my whole life. Coding will become obsolete.
@@cryptocontroller5585 Coding can of course become obsolete, so does every single profession you can imagine thanks to robotics and AI capabilities. I wonder how the banksters and politicians will take this change. Over the long run, financial systems will need to be redesigned, democracy as we know it will cease to excist and the definition of value will change. Losing a job will be the least of our problems. Also a layman who never coded will not have the necessary edge to direct a bunch of AIs where they fail to realize his vision. You need an innate understanding of what goes on under the hood, to judge the product these AIs produce. It is simply wishful thinking of those too lazy to learn some new skills and hoping that milk and honey will flow, once the stronger AIs are here.
Incredible video man. thank you for sharing step-by-step. would be cool to see a series of projects built with AutoGen. could make for some great content. Excited to see more experiments!
Darn it! 😮
I’m a software developer and I’m speechless!
Wow! Where was that when it was needed back then when I was still learning! 😅
❤
A development AI team sounds pretty neat to me as it bridges the gap between AI development (speed) and person development (creativity). I'm looking forward to custom UIs for this idea, you could seriously make a video game like program where you choose what each agent does in the team, what they look like and what they sound like. Then after you've created your characters you start talking to them. Your interactions result in the development of products and services.
Really looking forward to the fact that you can continue to update the AutoGen instruction, especially the multiplayer interaction with more examples, fantastic!
Honestly the notebooks like Google colab are awesome
Hey can you make another video as a add on to this one on how we can add multiple agents, create different agents, and how to make them all work together? I am brand new to coding and all this so I would really appreciate it. Love this video man you explained it in a way that even me, who has zero coding experience, was able to follow along. Thanks!
I second this!
This was nothing short of amazing. Thank you! Time to create my new company of AI workers :)
Today Autogen tomorrow to rule the world.... moar-har-har-har!
Thx for this Tutorial. Would love to see some more Advanced best practices.
Excellent level of explanation- most are either too simple to be useful or wizz through jargon at a bewildering rate 👏🏻 I’m interested in iPad instructions and creating more complex agents.
Next year? I think you meant to say Next WEEK! lol. Your instructions are Bang on! Keep it up. Thank you. Great work. Also quite entertaining at times : ) Super !
Can you please go into detail about the advanced AI agents? What role options are there? Use cases? How can we use the examples in the document? Which implications need to be there that the agents work well? Which problems do we usually face and how do we deal with them? So, basically, a lot of Python logic which is based on the Autogen Code. The audience is for non-techies
The best AUTOGEN tutorial ,nicely exlained
#tech - Have you tried customizing for using Opensource LLMs like Llama2 etc?
I love you for just trying to teach me how ❤
This is by far the most helpful video for absolute beginners. Subbed!
You’ve motivated me to break my gaming habit and tinker with AI agents instead. First idea is replicating functionality from your previous vid. Will try to have an AI create a Speech to Text tool that can cue to a specific keyword(s) within a RUclips video / Podcast. Often hard to find sound bytes within long audio files without extensive search. Edit: unless this step can be skipped with multi-modal advances, no conversion needed?
Anyone know where this type of noobie sharing discussion and ideas are held? Perhaps in the discord Wes mentions, but that may be more dev types rather than tinkering noobs like myself.
Look forward to your advance tutorial, hopefully not over my head!
Lol have the same type of idea. I want to be able to download a video then use a "sound" to text converter agent.
Then agent 2 to use this text and adjust it to a decent size before passing to the next agent.
agent 3 will read all of the information and make it a summary and try to extract all valuable information and then send to agent 4
agent 4 will take the adjusted content and try to make a final summary.
There's just so much stuff out there that i want to keep up with. But don't have the time to listen / watch and i think this could be a great way to start.
Good luck on your endavour fellow noobie.
good video as always. Love how you didn’t edit out your Weird Al distraction.😂
Tilde is [ ~ ] whereas the [ ` ] is a backtick among multiple other common names -- Pedantic dev
Pedantic appreciator here, I approve this message
Senior architect here. Message approved. I hate gaslighting devs
Not a dev but a pedant also appreciate
I like to call the ~ horizontal squiggle and i like to call the ` a flicky mc shtickens :P
This is very good - well done
Man! I love your videos, lots stuff to learn, to absorb, keep up the good work!!
Can’t wait for the prompt for this 😂. Very interesting though. Thanks Wes. I really appreciate your insights.
Great video, thank you!
For somebody who struggled with the config file like me... Here is a more straightforward way to pass Openai API key
from autogen import AssistantAgent, UserProxyAgent
assistant = AssistantAgent("assistant", llm_config={"api_key":"YOUR_OPEN_AI_API_KEY"})
user_proxy = UserProxyAgent ("user_proxy", code_execution_config={"work_dir": "coding"})
user_proxy.initiate_chat(assistant, message="plot a chart of NVDA and TESLA stock price change YTD.")
Fantastic. Thank you very much.
plot twist - most of Wes Roth's videos are made entirely with AI
as a large language model, I cannot confirm or deny this :)
I've been suspecting this for a while. He's the AGI is easing us into the revolution to help us realize it is not to be feared.
😳
I certainly hope so!
At this stage, it'd be a significant plot twist if not
Just wanted to to say thanks for the video.
Thought you would be excited about this Wes 😀
I'm very excited!
But there's TOO MUCH AI stuff coming out...
no time to sleep...
we need to pause AI for 6 months :)
@@WesRoth well, you have the tools to help you track, summarise and present the new stuff as it arrives. Lean more on AI to handle the work load, with you directing from what deeply interests you. You're doing an awesome job so far. 🙏👍
Nice thumbnail TBH I thought this was a gaming vid. PLEAse keep up the good work. I can see this having 10m views in 3 years and all your videos a 1m+
This is gold. Thank you sensei!
Thanks my guy
Thank you Wes!! Much Appreciated!!
I followed the tutorial but when trying to run it comes up with both Autogen not recognised- even though I pip installed pyautogen. And it says that the “config_list file ‘OAI_CONFIG_LIST’ does not exist”. What do I do next?
I ran into the same issue. Did you get any solutions?
Great and very well explained tutorial. I liked the step by step approach. I am running into rate limit issues with the GPT-4 model. How can I resolve them?
how can you change which GPT model autogen is using? Can you easily switch between 3.5 and 4.0?
What do we implement to switch from gpt 4 to gpt 3.5 turbo #tech
I think we´ll see companies whose main product is providing an agent for use in Autogen-like apps, fine-tuned or trained on their proprietary knowledge. And building those is the ai job market!
Very nice tutorial, but it's clear to me that at this point in time, cost is the major problem with this. Really looking forward to the rumored 20x price slashes in November!
Is this a sure thing? If so, I’ll be diving in at that time
#tech are there any talks of using non proprietary, free LLMs with autogen?
Thanks for sharing , i'm just looking for it and you video came out !
#tech Hey man, thanks for the great video! At 23:28 you introduce the critic with his own specific role description. Could you please share where you got this description from so we can take that specific example as inspiration and transfer it to different roles. Thanks again!
had a hell of a time installing autogen, after much back and forth with chat gpt I figured out I had the wrong version of python installed. It didn't work with the latest python version 3.12 had to downgrade to version 3.8 and it finally installed.
You were so great explaining everything! No jargon and just to the point, however, when you didn't show the snake game created by the group of agents I almost punched the monitor #tech
Sorry!
This video was really hard to make, I think I've skipped a bunch of steps.
I'm hoping to do another, more advanced video soon.
With the snake game, I think adding those bricks "bricked" the game lol.
It didn't start and at that point I was too tired to troubleshoot.
I think I almost punched the monitor myself :)
@@WesRoth I really appreciate what you do! Thanks 🫡
This is blowing my mind
Thank you very much 🙏
Coll brother...love your work and please do continue building more agents that can work together...love from Cape Town
My question is what's better this or chat dev
I think this WILL be a lot better over time. More resources and attention on this.
ChatDev is very cool, right out of the box, but my guess will be made obsolete over time.
Great video Wes! ... But ehmm.. I'd like to see the snake game made by the agents! 😄
This will use a huge amount of money, the name agent refers to automation, the basic and the minimum required feature is to at least Auto create the files, it's good to see them talking for a better code, so I think the best title for the video should be how to create a group chat AI, because that's the advanced version of your implementation
Suggestion: you can Implement function calling at least
This is great, I'll try this
Great video, cant wait to see how this evolves and the capabilities grow, i think this is the true future of ai. One thing i would like to know is how to use 3.5 instead of gpt4 as i dont have it atm. Are we likely to have a version where we arent paying per action in the distant future in your view?
Having a ton of agents doing a metric ton of trivial tasks and micro actions would add up significantly cost wise. Do you think it will get cheaper to the point where we have something similar to an ISP monthly unlimited use charge but fair usage/throttling in the early days but eventually progress to true unlimited as we now enjoy ISP bandwidths?
Yay agent npc's \o/
Wonderful, thank you! I think you will want to mask your API key in the video, or invalidate that key now that the video is out. I would not want to see people using your key and charging your account.
if I remember correctly you can give different agents different models - is that a way to bring down cost? Eg only one agent in the conversation uses gpt4? Or will every word the gpt4 agent reads count?
Just to note the youtube playback buttons are white, and white background screens blend them in at the bottom.
Hey man, don't forget to delete your keys after ;) Great content!
Amazing videos!! Could we create agents with Llama or other opensource LLMs??
Made by ai or not. I like the way he says monayy (4:22) please don’t stop 🙏
THIS IS AMAZING, now i wish i took python seriously. #tech 😄
Uncle Remus said Cool!
Supwr helpful, thank you. Be interesting to cover it doesn't do well at.
agreed. most of the fails were me trying to set stuff up.
hoping for a more advanced video soon, with more cool stuff (including it failing at tasks)
This would be very useful and I'd love to use it but I suspect like open interpreter which I've been using it's going to be insanely expensive running off gpt 4 api unfortunately.
I'm going to try running it offline, I've had llama code 2 running on 3090 at good speeds, so im hoping I can just host a server with api key from the same pc if that makes sense, Oi is awsome, but this is so much more!
What is the difference / benefit of using Autogen Vs LangChain (or even Langflow), or do they do different things? Can they be joined together?
Appreciate you!
Bro this is crazy
Does this work with gpt 3.5? Can we change the GPT version? I tried to and it doesn't work, but I'm a noob.
#tech The score keeping disappeared from a bout 15:25 into the video 🤔
Other than that, I really love your videos. Keep it up!
People dont want to believe that Python is the future. its is.. the popularity has been through the roof and its only going to grow. 5-10 years.
Could you please build a similar tutorial in wich ai agents get the data from a json file or database?
Thanks for the video, I have been following your content for months. I think it's a bit confusing to use IDLE or a text editor to edit the Python code when you already have Pycharm open. Why not just use Pycharm? This could be confusing for viewers new to coding.
@WesRoth You might want to disable that API key now. It got left in the video...
Make that both keys.
Thanks! Yup, all those are now gone.
#tech hi love this video but im running in to an error at the 'pip install pyautogen' step. I receive the following error: "ERROR: Could not build wheels for aiohttp, frozenlist, multidict, yarl, which is required to install pyproject.toml-based projects". Would appreciate your help if you know why.
same here, not really sure how to proceed
This video seems solved it for me ruclips.net/video/bLb6833hfRU/видео.html&ab_channel=SantoniGeek
same here
Great introduction! What is the cursor tool you are using?
Thank you for this well explained video. Is it possible to use the framework with Google Palm api ?
Great video and well explained! Thanks a lot. But I have big trouble to get some output from Visual Studio Code, do you have any ideas why?
How does this differ from just regular GPT4 with more manual intervention?
#tech Thank you for the video! Could you make a video on using Autogen to create a marketing company or a webscrapping company? Thanks again
Hi. Really nice job on this video. Thanks for sharing.
I have a question for you ... from your video, you ran your code using the code line: config_list = config_list_from_json(env_or_file="OAI_CONFIG_LIST").
I cannot make this work. I can replace this with my own hard coded config_list, but the error I get when using the code above is that OAI_CONFIG_LIST is not recognized.
Can you help with explaining why your code worked and recognized OAI_CONFIG_LIST without defining it first? Thanks!
the juicy part starts at 23:04
@wesroth - where does Azure OpenAI Services come in here? Can you use Autogen with Azure Openai services? or are they totally separate things that can't be used together?
There is an option to use Azure API. I haven't tried it yet, but planning to soon. Hopefully will have a video up on this at some point.
#tech How do I get specific agents to work in a hierarchy or structure of my choosing? For example; a CEO who deals with top level managers for Marketing, Development, Art, and then 2 agents below each of these who do not interact with the CEO.
ChatDev might be easier to do that out of the box:
ruclips.net/video/5Zj_zstLLP4/видео.html
With AutoGen, I think the command that customizes that is the 'GroupManager'. I think by default everyone is talking to everyone else.
More and more people are using and reporting on this, so we should see more tutorials soon, I'm still figuring stuff out myself.
@@WesRoth Maybe I'll just add the functionality into autogen if it's not there already.
@@JohnLewis-old Were you able to add that? If so, what process did you take?
I haven't figured it out yet, but my time trying was limited. @@Taytodude
My issues with these, is that they are very limited with the api token calls, can't experiment, can't use them freely.
#HELP Where do I get this complete code from? I'm having issues getting from one assistant to 3 or more. Any help would be greatly appreciated. I'm so close but so not there yet.
Do you share the code of the advanced agents demo somewhere? Thank you.
yup, all of it is on their github page. Check the link I have to natural20 that will have all the links details and all my tutorials in one place. (more coming soon!)
Thank you for curing my stupidity
Im completely soft brained when it comes to coding but id like to learn more now! Is there a way I could use autogenerated to set up some kind of e-commerce business? Im curious what actually creative ways people are already using this for lol
Could it be used to organize my OneNote inbox?
BRO! Your API KEY!!!!
Can you post your completed code? I'm so close.
dont you have to add OAI_CONFIG_LIST externally ?
pessimist : AGI is still long way to go.
me : prompt, you are wes roth. explain me about autogen, explain it in simple way, do not complicated like professor do. reply in RUclips video.
wes roth : I'm living LLM Explaining things in simple way. better than NLP professor.
I wonder what would happen if you assign one agent to every method in the code, or at least per file. Every one of them would know what their method does. In some bigger projects. Then main agent would assign necessary things to do for everyone on their own at their place.
Alright I keep following stro by step by can't get the provided code to work right. Any advice?
how to do this with azure provided keys