GREAT QUESTION. actually, the only decorators that are essential for crewAI to work are the ones from the tools (because the Langchain agents behind the scenes rely on them). that is why i only used those in the past video. all other decorators are there just to make your life "easier". these other decorators are part of the newer project structure of crewai, but they are not mandatory (for now). they are essentially just wrappers around your functions that run some extra code before and after your function. so, for example, the decorator @task does this when you add it to your function `func`: ``` def task(func): func.is_task = True tasks_order.append(func.__name__) return func ``` in other words, this decorator takes your method and adds the property `is_task` to true, and then adds it to an order list of tasks. as you can see, this will be useful to automatically load this method as a task if you are using the project structure recommended by CrewAI. but if you are not, then you can just create your tasks as we did in the previous video and add them to your Crew instance manually. the reason why i chose to show how this works without the decorators in the previous video is to show you crewAI in its simplest form. especially, because the framework is changing super fast, so it is essential to see what all these decorators and stuff actually boil down to. you can very well set up the crew from this video without the the decorators @task, @agent, @crew, etc.
Big shoutout to Alejandro AO - Software & AI for the incredible work you're doing! 🌟 As someone based in Nairobi, Kenya, accessing quality resources on software and AI can be a challenge, but your content has been an absolute game-changer for me. Your dedication to sharing knowledge and empowering others in the field of AI is truly commendable.
Outstanding!! You continue to produce some of the best hands on AI content on RUclips. Thank you and I really look forward to a follow up discussion on an optimized version, as insights from that will be very helpful... especially given the $$$ API calls for chatty and confused agent systems.
hey there! i'm glad you liked it! it is super useful. i hope more people used the Whispering extension. the developer really deserves recognition for it, especially for making it open source: github.com/braden-w/whispering
As ever congratulations for the growth of your channel, please keep on making content like this, we want full stack with crewai to automate HR processes please such a video like that thanks from Saudi Arabia
Hey AO, thanks for another great informative video. I got it to work with Google's Gemini. These OpenAI API calls could get expensive fast. Looking forward to your next video.
The title of the video got me hooked, the video clearly shows you're great at what you do so I wanted to ask, tasks like these are being automated using no code tools a lot nowadays within fraction of time, what's your opinion on those? Appreciate it
Hey, I loved your content, I am also experimenting with crewAI, chatDev and AutoGen. And what I am really curious about is- can we do what you just did using agents? Think of it like a HR agent. You just input a high level goal and let the HR agent populate the agents and tasks file. And once its done "hiring" the right guys, it calls on a custom tool that kickoff the secondary crew.
@Alejandro_ao this is the best video I have seen on crew-ai so far. Especially using the new file structure which has very limited docs. I am curious though on how the other agents/tasks have access to the .md files that you create during the intermediate steps to use as context.
Amazing tutorial. So we well explained and laid out. Question for you on CrewAI - is the process of picking a tool prescriptive, meaning the agent instructs which tool to pick? Or can the LLM pick automatically from a set of tools and decide what to use. I am just coming up to speed on CrewAI and wasn't sure. And your English is fantastic, by the way. First thought your accent sounded like you were from the USofA
wow incredible, thank you, great material, this is how to cook a crewai; What cost did this deployment of the crew generate for this demo, can you share it with us? And could you explain in detail how delegation works at a conceptual level with practical examples rather than at a technical level, and the aspect of how to control the flow at a technical level? I know you've already explained things, but it would be great to know the full detail. Thank you very much, and congratulations for your excellent work, good luck Compadre
@@alejandro_ao What I am expecting is, like PDF and web chat applications are getting very common nowadays, so if there will be a project which is unique and helpful too, that will be a great idea.
Hi Alejandro, can you please upload a tutorial of the Streamlit app for arxiv chatbot which includes scrapping of research papers from arxiv and we do question answering?
hey there! they offer a hosted service where they turn your crews into apis, but yeah that is paid. howerver, as far as i understand, it comes with much more than that as they actually personally help you set up your workflow with their experts
thanks! actually, WebBaseLoader already uses beautifulsoup, but facebook and instagram block scrappers like that. Browserless is probably the best alternative for scraping these kinds of tough sites, they get around most of these headaches (but they are a paid service :S): www.browserless.io/
@@alejandro_ao Mister, please check OS-world, if you havent already, is an OS designed to emulate human behaviour, this video tutorial + OS-world would be amazing
Hi Alejandro, could you explain to me how to add more qualifiers in yaml file, besides customary role, goal, backstory, but to identify tools, max_iter, or max_rpm? It it possible to pass that extra parameters via yaml files? If not, what is the way around it? Thanks!
Hello thank you so much for this valuable tutorial. I did the same steps but i got an error in Search tool " No module named 'langchain' howerver I installed the langchain package. Thank you again for all your efforts
Love and much appreciate the work/insight you share! been following you for a while on several accounts and have been meaning to ask: Are you against work for hire, contract work, or joining on? If so, very Interested in discussing Real Estate Use Case another dev and myself have been working on for a few months now. Could use some help pushing v1 across the finish line..have several clients in the pipeline just waiting for my call so they can pull trigger. Look forward to hearing from you either way!
I saw some other videos on CrewAI and nobody is talking about .yaml files like you have shown. What are yaml files and how do they differ from the .py files that others have shown. Even in the repos of CrewAI I did not see any yaml files
Ground level project. Lets talk hit me up. 1. CodeCraft Duel: Super Agent Showdown 2. Pixel Pioneers: Super Agent AI Clash 3. Digital Duel: LLM Super Agents Battle 4. Byte Battle Royale: Dueling LLM Agents 5. AI Code Clash: Super Agent Showdown 6. CodeCraft Combat: Super Agent Edition 7. Digital Duel: Super Agent AI Battle 8. Pixel Pioneers: LLM Super Agent Showdown 9. Byte Battle Royale: Super Agent AI Combat 10. AI Code Clash: Dueling Super Agents Edition
I find this is too complicated for a beginner who has no code experience. Please suggest what previous knowledge is needed to be able to build a crew like you have shown in the video? I am keen to learn how to create crews for my own needs so please give me a detailed answer
hey there, crewai is a very plug-and-play framework. you should be able to use it without much knowledge about what is going on behind the scenes. however, i do think that something that could be very useful to understand is how agents work, at least on a very high level. try checking out what are ReAct agents, that will shed some light on what is going on here. i will upload a video about this soon anyways
I would love to see a local example. I have tried to get this and other examples to run locally but I have not had any luck. I have followed the example on the crewai page and other videos but cant figure out why the tools wont work. Flip it back over to "open"AI and it runs but burns tokens too fast.
@@alejandro_ao perfect. exactly what i've stumbled upon and have been playing with for a couple of days now. I messed around with Flask at first cause i've used it in the past. But streamlit is dead simple to use. quite happy with it.
Traceback (most recent call last): File "E:\Automation\crewAI-examples\markdown_validator\Main.py", line 2, in from crewai import Agent, Task File "C:\Users\HIG Ajay\AppData\Local\Programs\Python\Python310\lib\site-packages\crewai\__init__.py", line 3, in from crewai.crew import Crew File "C:\Users\HIG Ajay\AppData\Local\Programs\Python\Python310\lib\site-packages\crewai\crew.py", line 19, in from crewai.agent import Agent File "C:\Users\HIG Ajay\AppData\Local\Programs\Python\Python310\lib\site-packages\crewai\agent.py", line 6, in from crewai import Agent, Task, Crew ImportError: cannot import name 'Agent' from partially initialized module 'crewai' (most likely due to a circular import) (C:\Users\HIG Ajay\AppData\Local\Programs\Python\Python310\lib\site-packages\crewai\__init__.py) how to fix this error ?
hey there, there’s a waitlist for their hosted platform, crewai+. but you can use the package as i do here just by installing it with “pip install crewai”
I need some one who can help me to create the crew shown in this video so that I can use it for my IG marketing. I would need some customisation for my requirements. Please contact me if you can do the needful
![Coheed and Cambria - Searching for Tomorrow [Official Video]](http://i.ytimg.com/vi/hDsYGLbmadg/mqdefault.jpg)
why is it that in this video you were putting the decorators whereas in ur 1st crewai video, you didnt. I'm confused
GREAT QUESTION. actually, the only decorators that are essential for crewAI to work are the ones from the tools (because the Langchain agents behind the scenes rely on them). that is why i only used those in the past video. all other decorators are there just to make your life "easier".
these other decorators are part of the newer project structure of crewai, but they are not mandatory (for now). they are essentially just wrappers around your functions that run some extra code before and after your function. so, for example, the decorator @task does this when you add it to your function `func`:
```
def task(func):
func.is_task = True
tasks_order.append(func.__name__)
return func
```
in other words, this decorator takes your method and adds the property `is_task` to true, and then adds it to an order list of tasks. as you can see, this will be useful to automatically load this method as a task if you are using the project structure recommended by CrewAI. but if you are not, then you can just create your tasks as we did in the previous video and add them to your Crew instance manually.
the reason why i chose to show how this works without the decorators in the previous video is to show you crewAI in its simplest form. especially, because the framework is changing super fast, so it is essential to see what all these decorators and stuff actually boil down to. you can very well set up the crew from this video without the the decorators @task, @agent, @crew, etc.
Big shoutout to Alejandro AO - Software & AI for the incredible work you're doing! 🌟 As someone based in Nairobi, Kenya, accessing quality resources on software and AI can be a challenge, but your content has been an absolute game-changer for me. Your dedication to sharing knowledge and empowering others in the field of AI is truly commendable.
it's my honor. i only provide some lessons, it's up to you to use this to build super cool stuff that makes an impact in the world, you can do this 🚀
@@alejandro_ao Absolutely
it's amazing how high the quality of your materials and the knowledge they convey is
best video ever i have seen about crew , congrats i hope you make more about how to make a ui interface of this . thanks you so much
i really appreciate it
I’ve been very interested in AI Agent swarms the last few weeks. Thanks Alejandro for the video
thanks to you, expect much more agentic topics coming soon!
Outstanding!! You continue to produce some of the best hands on AI content on RUclips. Thank you and I really look forward to a follow up discussion on an optimized version, as insights from that will be very helpful... especially given the $$$ API calls for chatty and confused agent systems.
thank you! absolutely, i will be showing how to use these in production (albeit a rudimentary version) as well!
Bro that ChatWithOurDocs improvement was LENGEDARY!! Thank you for introducing me to "Whispering" 💜🙌
hey there! i'm glad you liked it! it is super useful. i hope more people used the Whispering extension. the developer really deserves recognition for it, especially for making it open source: github.com/braden-w/whispering
哇，再次感謝您為我們帶來的高品質教學視頻，絕對為我帶來實質的幫助！
不客氣！ 祝你有美好的一天！
As ever congratulations for the growth of your channel, please keep on making content like this, we want full stack with crewai to automate HR processes please such a video like that thanks from Saudi Arabia
Great video. Looking forward to the next one
Hey AO, thanks for another great informative video. I got it to work with Google's Gemini. These OpenAI API calls could get expensive fast. Looking forward to your next video.
How can you make it using Google Gemini api key
wow, impressive use of CrewAI! :D
what would you like to see next?
The title of the video got me hooked, the video clearly shows you're great at what you do so I wanted to ask, tasks like these are being automated using no code tools a lot nowadays within fraction of time, what's your opinion on those? Appreciate it
Hey, I loved your content, I am also experimenting with crewAI, chatDev and AutoGen. And what I am really curious about is- can we do what you just did using agents? Think of it like a HR agent. You just input a high level goal and let the HR agent populate the agents and tasks file. And once its done "hiring" the right guys, it calls on a custom tool that kickoff the secondary crew.
This is amazing Alejandro, thank you! Just wondering how many tokens the test output you provided would cost roughly?
very nice video, thanks you, liked and suscribed, which program do you use for record the video? thanks you so much
welcome to the channel :) i believe i was using Screen Studio for some takes and OBS for others :)
@Alejandro_ao this is the best video I have seen on crew-ai so far. Especially using the new file structure which has very limited docs. I am curious though on how the other agents/tasks have access to the .md files that you create during the intermediate steps to use as context.
Glad it was helpful! that is a possibility indeed, using those md files will make for a much more complex and complete application
Amazing tutorial. So we well explained and laid out. Question for you on CrewAI - is the process of picking a tool prescriptive, meaning the agent instructs which tool to pick? Or can the LLM pick automatically from a set of tools and decide what to use. I am just coming up to speed on CrewAI and wasn't sure. And your English is fantastic, by the way. First thought your accent sounded like you were from the USofA
Franchement un grand bravo pour le travail tu maitrises :)
grand merci !
Explore the cutting-edge of AI with us! Subscribe for insightful, easy-to-follow updates.
patiently waiting for the groq api video 🙏
This is fascinating! Thank you for sharing this!
Great video wow i wish i would be done something like this in the future
Keep posting ❤
Amazing value in your video !!! Thank you for that !! 🥳
thank you for being here!
Epic Content 🙏🏼🙌🏼
you are epic
Awwwsome work man!
can you please share the link for the notes for this video that you used ?
Absolutely love your content. What auto completion engine are you using for VSCode/cursor ide?
that would be gh copilot!
wow incredible, thank you, great material, this is how to cook a crewai; What cost did this deployment of the crew generate for this demo, can you share it with us? And could you explain in detail how delegation works at a conceptual level with practical examples rather than at a technical level, and the aspect of how to control the flow at a technical level? I know you've already explained things, but it would be great to know the full detail. Thank you very much, and congratulations for your excellent work, good luck Compadre
What I was looking for!
let me know what you would like to see next. i'm thinking guis
Excellent work. looking for more sample use cases on crewai to learn more. Thanks.
coming up :) what would you like to see?
@@alejandro_ao Maybe smth like: CrewAi(RAG + WebSearch) => excel as repo + ppt :)
loved the video, thanks for sharing
Thanks for watching!
Thank you!! any plans for a CrewAI + Streamlit video?🙂
Very interesting buddy ❤
Please make some more unique llm project with streamlit/web app. Your previous projects were really awesome.
I agree. I have used Streamlit UI and it’s very clean. A good tutorial would be great on creating a UI
thanks for the feedback, man! yeah, i will be implementing more of that! 🫡 let me know if there are any implementations in particular that you expect!
Yeah, that's right. Waiting for more cool projects
@@alejandro_ao What I am expecting is, like PDF and web chat applications are getting very common nowadays, so if there will be a project which is unique and helpful too, that will be a great idea.
Hi Alejandro, can you please upload a tutorial of the Streamlit app for arxiv chatbot which includes scrapping of research papers from arxiv and we do question answering?
Crewai can expose an api for free? Or that feature is paid? Thanks you so much
hey there! they offer a hosted service where they turn your crews into apis, but yeah that is paid. howerver, as far as i understand, it comes with much more than that as they actually personally help you set up your workflow with their experts
Awesome video! 🙏👍 Can we set a user agent in Beautiful Soup perhaps? To fix the scraping issues?
thanks! actually, WebBaseLoader already uses beautifulsoup, but facebook and instagram block scrappers like that. Browserless is probably the best alternative for scraping these kinds of tough sites, they get around most of these headaches (but they are a paid service :S): www.browserless.io/
@@alejandro_ao Thanks for the tip! :)
@@alejandro_ao Mister, please check OS-world, if you havent already, is an OS designed to emulate human behaviour, this video tutorial + OS-world would be amazing
Hi Alejandro, could you explain to me how to add more qualifiers in yaml file, besides customary role, goal, backstory, but to identify tools, max_iter, or max_rpm? It it possible to pass that extra parameters via yaml files? If not, what is the way around it? Thanks!
continue pls with this content
coming up this week :)
i like you learning and implementing style can you guide how i improve my learning speed in this dynamic tech world
🙃🙃
Hello thank you so much for this valuable tutorial. I did the same steps but i got an error in Search tool " No module named 'langchain' howerver I installed the langchain package.
Thank you again for all your efforts
ooh man i need to have some good system to even consider this!
what do you mean? we are running all of this on the cloud, no need for a super big computer
@@alejandro_ao i thought i could run it my local systen
@@alejandro_ao .. need supported macOS version though... tried running it on my 10-year old Apple MacBook... "not supported"
Hello mister, is there a way to interact with Instagram interphase? To make this fully autonomous
awesome!
Thank you! Cheers!
Love and much appreciate the work/insight you share! been following you for a while on several accounts and have been meaning to ask: Are you against work for hire, contract work, or joining on? If so, very Interested in discussing Real Estate Use Case another dev and myself have been working on for a few months now. Could use some help pushing v1 across the finish line..have several clients in the pipeline just waiting for my call so they can pull trigger. Look forward to hearing from you either way!
I saw some other videos on CrewAI and nobody is talking about .yaml files like you have shown. What are yaml files and how do they differ from the .py files that others have shown. Even in the repos of CrewAI I did not see any yaml files
i wanted to make this as approachable as possible. i am glad that was useful to you!
How can I use it with groq instead? do I just set the key in the enviroment or do I have to change something somewhere else?
Is it possible to connect to Azure Open AI instead of Open AI?
Ground level project. Lets talk hit me up.
1. CodeCraft Duel: Super Agent Showdown
2. Pixel Pioneers: Super Agent AI Clash
3. Digital Duel: LLM Super Agents Battle
4. Byte Battle Royale: Dueling LLM Agents
5. AI Code Clash: Super Agent Showdown
6. CodeCraft Combat: Super Agent Edition
7. Digital Duel: Super Agent AI Battle
8. Pixel Pioneers: LLM Super Agent Showdown
9. Byte Battle Royale: Super Agent AI Combat
10. AI Code Clash: Dueling Super Agents Edition
I would be really really great full if you could make tutorial on multiple pages automated data dashboard with automated chat bot….
I find this is too complicated for a beginner who has no code experience. Please suggest what previous knowledge is needed to be able to build a crew like you have shown in the video? I am keen to learn how to create crews for my own needs so please give me a detailed answer
hey there, crewai is a very plug-and-play framework. you should be able to use it without much knowledge about what is going on behind the scenes. however, i do think that something that could be very useful to understand is how agents work, at least on a very high level. try checking out what are ReAct agents, that will shed some light on what is going on here. i will upload a video about this soon anyways
top top top
Hi! How can one create a Tool for the agent to read pdf? Thank you!
best one
I would love to see a local example. I have tried to get this and other examples to run locally but I have not had any luck. I have followed the example on the crewai page and other videos but cant figure out why the tools wont work. Flip it back over to "open"AI and it runs but burns tokens too fast.
i'll start working on a video about this :)
@@alejandro_ao Awesome! Thank you!
What gui would you recommend for this use case?
streamlit for sure. i have video coming up this week about this
@@alejandro_ao perfect. exactly what i've stumbled upon and have been playing with for a couple of days now. I messed around with Flask at first cause i've used it in the past. But streamlit is dead simple to use. quite happy with it.
I'm getting "zsh: command not found: crewai" when I write in terminal "crewai create [NAME]", any advice?
Put "conda" first if you're using it
Where is the conda install link?
Can i get any new idea where i can automat something.I mean the scenario where i can make my own project using crew.please suggest me something
How can I connect with you?
Thank
you
how did you install the sdk?
it comes with pip install crewai
Is it possible tô create the rules, tasks in onother language than English?
absolutely! but you will have to add your own translation of a i18n file with the base prompts... i will show how to do that very soon
Traceback (most recent call last):
File "E:\Automation\crewAI-examples\markdown_validator\Main.py", line 2, in
from crewai import Agent, Task
File "C:\Users\HIG Ajay\AppData\Local\Programs\Python\Python310\lib\site-packages\crewai\__init__.py", line 3, in
from crewai.crew import Crew
File "C:\Users\HIG Ajay\AppData\Local\Programs\Python\Python310\lib\site-packages\crewai\crew.py", line 19, in
from crewai.agent import Agent
File "C:\Users\HIG Ajay\AppData\Local\Programs\Python\Python310\lib\site-packages\crewai\agent.py", line 6, in
from crewai import Agent, Task, Crew
ImportError: cannot import name 'Agent' from partially initialized module 'crewai' (most likely due to a circular import) (C:\Users\HIG Ajay\AppData\Local\Programs\Python\Python310\lib\site-packages\crewai\__init__.py)
how to fix this error
?
Use opensource LLM or provide the way to use opensource LLM
Yep agree, privacy will be the main issue with this oaid LLM
Can I hire you to set this up for us at our company ?
Hey there, I will start doing consulting soon. Send me a description of your situation here: hello@alejandro-ao.com
CrewAi is waitlisted for new accounts. So i hope soon this changes.
hey there, there’s a waitlist for their hosted platform, crewai+. but you can use the package as i do here just by installing it with “pip install crewai”
Fantastic Video now, please go and change your API keys so people can't rip you off ❤
i live in constant fear that someday i will forget to disable them before publishing 🥲
Google serper api isnt working after all.I used the serpapi for that and it wasnt working correct if i was returning the correct string
The girl name zara dar is an AI which u created, isnt it!! Mm ???😂
hah i reckon we have the same charm
I need some one who can help me to create the crew shown in this video so that I can use it for my IG marketing. I would need some customisation for my requirements. Please contact me if you can do the needful