Very nice and very explanatory. I wonder whether if you can modify the code making AgentManager it self a MasterAgent (a MasterRouterAgent leaded by a LLM) that, based on the user inputs, chooses the next suitable flow from a given flows (or respond to the used that it is unable to do the given task since it is not supported with the needed flow YET), Then: MasterRouterAgent chooses one from described groups (each group includes tasks that are belong or classified under same domain) to route to: Group1SecondaryRouterAgent Group2SecondaryRouterAgent Group3SecondaryRouterAgent . . etc Where each flow that is available for MasterRouterAgent has SecondaryRouterAgent that redirect the task to the suitable flow from grouped tasks specific flows where each flow consists of (Task1Group1ExecuterAgent->Task1Group1ValidatorAgent) i.e.: SecondaryRouterAgent choose from flows that are grouped under one group of tasks (Domain Tasks) to route to the suitable flow Group1RouterAgent selects from described flows: (Task1Group1ExecuterAgent->Task1Group1ValidatorAgent) (Task2Group1ExecuterAgent->Task2Group1ValidatorAgent) (Task3Group1ExecuterAgent->Task3Group1ValidatorAgent) Group2RouterAgent selects from described flows: (Task1Group2ExecuterAgent->Task1Group2ValidatorAgent) (Task2Group2ExecuterAgent->Task2Group2ValidatorAgent) (Task3Group2ExecuterAgent->Task3Group2ValidatorAgent) . . etc I believe this tree of flows can help in making small LLMs able to give acceptable results, since each LLM will be responsible on only one specific simple task-or-tool. Also the classification under domain specific branch (like, files tasks (create-read-append), web tasks(search-crawl-extract), database tasks(create-read-update-find-delete), ...etc. would help in making adding more groups and tasks specific agents easy since your code depends on classes (O.O.P.) May be such target can be done by asking for help from one of the clever powerful LLMs to write this code for us. Unfortunately, I am a beginner in python and I'm still learning how to write code. Anyway, thanks for the good and distinct content. 🌹🌹🌹
Hello everything is fine? I would like to ask you to see this voice generation model, text to speech, if possible, please make a video explaining it better! ruclips.net/video/o4NslS7figE/видео.html
![Ice On My Teeth - ATEEZ エイティーズ 에이티즈 [Music Bank] | KBS WORLD TV 241115](http://i.ytimg.com/vi/c-n92OgbbwA/mqdefault.jpg)
Please provide projects which are totally free of cost , use open source more often
I'll make sure to feature more free and open-source projects in future videos.
@AIAnytime thanks
Very nice and very explanatory.
I wonder whether if you can modify the code making AgentManager it self a MasterAgent (a MasterRouterAgent leaded by a LLM) that, based on the user inputs, chooses the next suitable flow from a given flows (or respond to the used that it is unable to do the given task since it is not supported with the needed flow YET),
Then:
MasterRouterAgent chooses one from described groups (each group includes tasks that are belong or classified under same domain) to route to:
Group1SecondaryRouterAgent
Group2SecondaryRouterAgent
Group3SecondaryRouterAgent
.
.
etc
Where each flow that is available for MasterRouterAgent has SecondaryRouterAgent that redirect the task to the suitable flow from grouped tasks specific flows where each flow consists of (Task1Group1ExecuterAgent->Task1Group1ValidatorAgent)
i.e.: SecondaryRouterAgent choose from flows that are grouped under one group of tasks (Domain Tasks) to route to the suitable flow
Group1RouterAgent selects from described flows:
(Task1Group1ExecuterAgent->Task1Group1ValidatorAgent)
(Task2Group1ExecuterAgent->Task2Group1ValidatorAgent)
(Task3Group1ExecuterAgent->Task3Group1ValidatorAgent)
Group2RouterAgent selects from described flows:
(Task1Group2ExecuterAgent->Task1Group2ValidatorAgent)
(Task2Group2ExecuterAgent->Task2Group2ValidatorAgent)
(Task3Group2ExecuterAgent->Task3Group2ValidatorAgent)
.
.
etc
I believe this tree of flows can help in making small LLMs able to give acceptable results, since each LLM will be responsible on only one specific simple task-or-tool. Also the classification under domain specific branch (like, files tasks (create-read-append), web tasks(search-crawl-extract), database tasks(create-read-update-find-delete), ...etc. would help in making adding more groups and tasks specific agents easy since your code depends on classes (O.O.P.)
May be such target can be done by asking for help from one of the clever powerful LLMs to write this code for us.
Unfortunately, I am a beginner in python and I'm still learning how to write code.
Anyway, thanks for the good and distinct content.
🌹🌹🌹
That's a really interesting idea! I'll think about how to implement that in future videos. Thanks for the suggestion!
Very informative bro, Thankyou.
Thanks for watching the video
Great tutorial
Glad you think so!
can add how to use code execution with python REpl custom tool and E2B sandbox
Yeah.... We discussed this in our group. One of the guys will be making a PR on GitHub shortly. Will update
Excellent!!
Thank you! Cheers!
Thanks:)
Welcome!
Can you build a chatbot clone of yourself to whom we can ask questions?
Great idea. Let me put this on the roadmap.
great
Thanks
first comment
Hello everything is fine? I would like to ask you to see this voice generation model, text to speech, if possible, please make a video explaining it better!
ruclips.net/video/o4NslS7figE/видео.html