Stunning. I've been using specific coding directives in the "Customize ChatGPT" feature of the basic interface to help me code my first react project and I'm amazed at how well it works. This Code Analysis Agency takes that concept to a whole new level with expanded and stricter directive enforcements. I want one. :)
Thank you for following me! Glad you find my work useful. Capabilities are definitely gonna get even crazier when OpenAI finally updates Assistants API.
Congratulations on the framework. Any possibility of having a graphical construction interface? A flowbuilder for those who don't understand code? a $9 monthly subscription to have flowbuilder?
Hey, just answered. Few agents that are fine tuned well are a lot more effective than 50 that are barely functioning correctly. Always start with as few agents as possible.
Good video! My question is why you are not using way bigger agent structures with 50 agents working together or so? Would that even be structurally possible? Ever tried it? Like a really long chain of scenarios that are worked through. I hope this question is not too abstract.
This might be a great idea on a scaled requirement but it doesn't seem economical for smaller projects/teams - assuming the time it would take to create an effective chain of 50 fine-tuned agents, and providing specific directives to each, would take more time than manually doing the task yourself. It's getting the chain to do everything you ask every time the same way that is the hard part. The less links in the chain the less chance it has for failure at some point where optimization for the task at hand is maximized. Or I might be totally misunderstanding this implementation or your question. My mind is just so blown by what is possible now that I don't know what to believe is reality any more. lol
@@Paul_Marek No that's exactly what my question was lol. I'm not sure because each agent would be able to decide when to use specific tools so in theory if the system is tuned well we should be able to do a specific "category" of work with such a system right? Idk, it would be an expensive experiment this is why I'm asking first but I guess this might be a situation where one just needs to try and see
Hey, I always recommend starting with as few agents as possible, and expending as necessary, once all previous agents are functioning reliably. If you start with 50 there is no way to debug the system and fine tune any single agent.
If you are going to release new options in your OA website you need info tutorials because I see alot of errors when trying to use it, when providing products you need info videos to support it.
Hello mate! I'm starting to use the framework but I'm facing some issues to do some things that should be easy (at least in my viewpoint). I did a lot of search and tried to find ways to connect my agents to S3, mongoDB, salesforce and other apis/dbs but completely without success.... Could you please record a video explaining how to create a agent to connect to these tools, please? Like the comment if you agree with me and want to see this kind of solution
![โรงเรียนดัดสันดาน นักเรียนเถื่อน..!! [โรงเรียนไดเม่]](http://i.ytimg.com/vi/vRKPeXn4Fl4/mqdefault.jpg)
I will answer as many comments as I can tomorrow. For now, stay tuned in our discord! ⚡(link in description) ☝
What an incredible automated workflow! 👏
Stunning.
I've been using specific coding directives in the "Customize ChatGPT" feature of the basic interface to help me code my first react project and I'm amazed at how well it works. This Code Analysis Agency takes that concept to a whole new level with expanded and stricter directive enforcements.
I want one. :)
Running it on our projects already, and there are a lot of ways to improve it even further! I might release another video later once we do that.
So pleased I found your channel! Enjoy your content, method of delivery, instructional value and thank you for promoting AI tools in a positive light.
Thanks! Glad it helps
this solution is so good it went from 3000 to 3250 in a few minutes !
Yeah, haha. You can sell it for a lot more, if you tailor it for a specific company)
I'd love to see a video on building an agency that generates code, fixes issues, and then commits! Think we could see you go through that process?
Incredible work as always and the capabilities keep on growing, Again and as always, thank you for making everything open source and being so active
Thank you for following me! Glad you find my work useful. Capabilities are definitely gonna get even crazier when OpenAI finally updates Assistants API.
Congratulations on the framework.
Any possibility of having a graphical construction interface? A flowbuilder for those who don't understand code? a $9 monthly subscription to have flowbuilder?
Another banger from the legend himself. Keep up the great work man! Really appreciate it 🙏🏾
Thanks!
Excellent framework, excellent tutorials 🎉
You answered every single comment except mine :D
Hey, just answered. Few agents that are fine tuned well are a lot more effective than 50 that are barely functioning correctly. Always start with as few agents as possible.
Good video! My question is why you are not using way bigger agent structures with 50 agents working together or so? Would that even be structurally possible? Ever tried it? Like a really long chain of scenarios that are worked through. I hope this question is not too abstract.
This might be a great idea on a scaled requirement but it doesn't seem economical for smaller projects/teams - assuming the time it would take to create an effective chain of 50 fine-tuned agents, and providing specific directives to each, would take more time than manually doing the task yourself. It's getting the chain to do everything you ask every time the same way that is the hard part. The less links in the chain the less chance it has for failure at some point where optimization for the task at hand is maximized. Or I might be totally misunderstanding this implementation or your question. My mind is just so blown by what is possible now that I don't know what to believe is reality any more. lol
@@Paul_Marek No that's exactly what my question was lol. I'm not sure because each agent would be able to decide when to use specific tools so in theory if the system is tuned well we should be able to do a specific "category" of work with such a system right? Idk, it would be an expensive experiment this is why I'm asking first but I guess this might be a situation where one just needs to try and see
@vrsen Would be cool if you could give some insight, I'm really struggling to understand what might prevent it.
Hey, I always recommend starting with as few agents as possible, and expending as necessary, once all previous agents are functioning reliably. If you start with 50 there is no way to debug the system and fine tune any single agent.
@@vrsen Thanks for the answer! But have you ever tried expanding to a high number? Because theoretically a real real lot should be possible already?
If you are going to release new options in your OA website you need info tutorials because I see alot of errors when trying to use it, when providing products you need info videos to support it.
Yeah I know. I am working on this
What the heck is up with the oa website i havent been able to login for days , and i have tried many browsers, something is wrong with it.@@vrsen
Thanks for more great work!
Hello mate! I'm starting to use the framework but I'm facing some issues to do some things that should be easy (at least in my viewpoint).
I did a lot of search and tried to find ways to connect my agents to S3, mongoDB, salesforce and other apis/dbs but completely without success.... Could you please record a video explaining how to create a agent to connect to these tools, please?
Like the comment if you agree with me and want to see this kind of solution
Bro how long until a Generalist Agency which is basically baby AGI? Genesis Agency is pretty close.
I will continue improving this agency after I finish documentation and multi agency communication. Then, things are gonna get interesting:)
Awesome as usual
Hi Vrsen, does it work only only with open ai api key?
Yes, only openai models.
Please consider increasing the font size so it is readable by your viewers.
Yes, my bad. Will do so on the next one
I am on windows, sudo does not run. Can someone please help.
Cost to higher for a project?
Depends, you can reach out at vrsen.ai
include claude sonnet 3.5 along with openai
🧠