CrewAI Flows Crash Course

Рет қаралды 14,886

Күн бұрын

Пікірлер: 48

@bhancock_ai 2 ай бұрын

Hey guys! I made some simple but much needed improvements to the way we write book chapters in parallel. To see these improvements and updates, please checkout the source code in the description. Also, thank you to everyone who has pointed out issues and improvements that I can make to the source code. I believe I've addressed all the issues and added in all the fixes mentioned so far! Y'all are awesome! Please let me know if you find anything else or have any other questions!

@robsucher9419 10 күн бұрын

Brandon - yours is one of the best videos that I've seen on CrewAI flows!

@ophelia6207 19 күн бұрын

This was such a great tutorial, thanks so much Brandon! Loving your content.

@RobShocks 2 ай бұрын

Brandon just posted, hold all my calls, boil the kettle and fire up your code editor!! Looking forward to this one

@bhancock_ai 2 ай бұрын

Love it! Can’t wait to hear what you think about flows!

@raghavbhandari6983 2 ай бұрын

Most easiest and structured way i have seen of the internet to learn CREW I really hope if u will be able to make long video on Langgraph with some real world project 🙌

@switzerlandbase 2 ай бұрын

Yes, Brandon`s approach is top notch, can wait to work on this

@bhancock_ai 2 ай бұрын

Thanks man!! We made so many improvements to flows that I had to record this like 3 times for you guys! I think y’all are going to love it!

@wahwahwally1 2 ай бұрын

awesome video! I'm at the 6-month mark with python now and you have inspired me from the beginning, ty!

@bhancock_ai 2 ай бұрын

That's awesome man! Don't give up! Keep going! Out of curiosity, what is your end goal? Launching your own apps? Landing a job?

@wahwahwally1 2 ай бұрын

@@bhancock_ai thanks! Right now i work with a lot of excel files so I started learning it to speed up my productivity. Now i can leave work 2 hours early. lol hopefully it will land me a data analyst job one day. I've been keeping on eye on RAG AI and you're my go-to currently.

@idaelpineiro8339 2 ай бұрын

Hi Brandon, Great to be back from vacation and reconnect with the group. I'm excited about the new CrewAI Flows crash course! Hoping it can help me automate my Mail2Sheet email extraction or email scraper system and leverage CrewAI's full potential. Cheers!

@aladinmovies 2 ай бұрын

Amazing detailed tutorial! Thanks Brandon!

@adammyers6540 Ай бұрын

Incredible information for the less skilled. Thank you Sir!

@TomPooleMS 2 ай бұрын

Thanks Brandon - you are the best!

@giladbar-ilan1472 2 ай бұрын

again thanks! btw, when running the book, i would expect in the plot to show that we have multiply chapters. this must be some UX improvement here

@thefutureisbright 2 ай бұрын

Excellent tutorial Brandon.

@WillianSilvanoMaria 2 ай бұрын

Thank you for this awesome video!

@Prathameshhh_ 2 ай бұрын

Sir, please create a video on interactive chatbot with crew ai having memory and being interactive throughout the conversation .

@giladbar-ilan1472 2 ай бұрын

Thanks for the detailed example. This is Great. How can i have a fixed chapter list for a book? I want each book to be with same structure, can u define the list of chaptters as a fixed value instead of using a crew to generate that?

@bhancock_ai 2 ай бұрын

Awesome question! The simplest solution would be to update the chapter outline generator crew to say please generate a chapter outline in this format: [Insert Format Here] ---- Or, you could add the chapter list as a fixed value in the `state` of the flow. I hope that helps! Please let me know if you have any followup questions!

@giladbar-ilan1472 2 ай бұрын

@@bhancock_ai so what you are suggesting is to override the outline_book_crew and set it as fixed values like this: chapters=[1. intro 2.{} 3.{}... X.summary]? did i understood you correctly?

@peddibharath7016 Ай бұрын

very easy to understand tutorial

@manuelbradovent3562 2 ай бұрын

Excellent! Looking forward to improve the flow. Just found that py312 works and runs flow nicely while 3.11 needs asyncio wrapper the kickoff?

@bhancock_ai 2 ай бұрын

Hey! We actually did a last minute improvement on flows to make them not async so you don't have to wrap them in an asyncio wrapper. If you update to ^0.76, you won't have to worry about it anymore. Please let me know if you have any other questions!

@manuelbradovent3562 2 ай бұрын

@@bhancock_ai I ran into an issue , while crewai flow kickoff works perfectly , the crewai flow plot wont and raises below error. I have tried to update all dependencies and will look further how to resolve it. In case you know any workarounds would appreciate. Error: Caused by: program not found An error occurred while plotting the flow: Command '['uv', 'run', 'plot_flow']' returned non-zero exit status

@mrknow-it-all7847 2 ай бұрын

Brandon, just wondering. What if you want to use something like a vector store as the repository for your book. Would you just change the researcher to use the vector store instead of the Internet?

@bhancock_ai 2 ай бұрын

Exactly! You would just give your Researcher a different tool. Instead of a search the internet tool, you'd give them a query a vector store tool. Good luck! Let me know if you need help with anything else!

@sola1356 2 ай бұрын

Hi Brandon, I loved the video-it was really well done! I’ve been using CrewAI for a while, but I occasionally run into exceptions that break the entire crew, and I haven’t found many resources on handling this. Do you have any suggestions on managing or troubleshooting these issues?

@bhancock_ai 2 ай бұрын

Hey! Do you have any examples that you could share with me so I can see what you're talking about?

@sola1356 2 ай бұрын

@@bhancock_ai Appreciate your reply:) In some cases, if the LLM cannot determine a final answer and reaches the iteration limit, an error message-'Agent stopped due to iteration limit or time limit'-may occur. To address this, I currently check the results once the crew has completed (e.g., ensuring data is updated in the database). If the expected updates aren't present, will rerun the crew. However, I'm uncertain if this approach is the most effective. Are there more effective ways to capture specific task errors, halt the crew, and resume from the exact task where the error occurred?

@anilshinde8025 2 ай бұрын

Very interesting and informative.

@ricardoquiroga4891 2 ай бұрын

Awesome! One question: Why do you use from langchain_openai import ChatOpenAI instead of from crewai import LLM ?

@bhancock_ai 2 ай бұрын

I'm a dummy. That's why lol! It should have been this one: `crewai import LLM` Great call! I'm going to do a followup more advanced crewai flows tutorial soon and I'll be sure to use the right one in the next video!

@Terresa_P 2 ай бұрын

Thank you, Brandon, for this fantastic tutorial! I’m curious if this augmented CrewAI still lacks the Cycles feature that LangGraph is known for, which loops back to the previous node for self-reflection?

@bhancock_ai 2 ай бұрын

hey! I’m so glad you brought that up! You can definitely do cycles for self reflection with flows! there is an example over on our website where you can set up a feedback loop with human in the middle. Here is the link to the example. github.com/crewAIInc/crewAI-examples/tree/main/lead-score-flow Also, you bring up a really good point about auto self reflection so I will go update and add an example to the docs right now!

@bhancock_ai 2 ай бұрын

Hey! I just created a new example in the crewAI docs and my next KZbin deep dive into flows with cycles. Here's the link in the mean time: github.com/crewAIInc/crewAI-examples/tree/main/self_evaluation_loop_flow Thank you for calling this out!

@faizantayyab8222 Ай бұрын

Is it normal for the Flow or CrewaI to take long time to initiate. What factors are to be considered in terms of performance. Is Crew AI slower then langgraph?

@mr.mentor699 2 ай бұрын

Awesome sir

@trohitattraya2350 15 күн бұрын

I'm tried using flows, but is there a way we can pass inputs to our flow as I'm not able to

@augmentos Ай бұрын

this is so not what I thought Crew AI was. Its basically programming lol you also did make a minor error at the start (I don't personally love the verbiage of 'flow' 'start' its confusing and I would have thought they would be named (if those names) inversesly. But you mentioned at one point (then self corrected) a start starts a flow. I'd have at the start explained the simple concept that a flow can have many starts, like you have a drive through a city and there will be many red lights on crossings with 'starts' or something to make it make sense. my .2cents. Good video.

@thornesmith1277 Ай бұрын

Ok, I'm feeling kind of stupid, like I can't see the mustard in the frig because It is right in front of me... But How does it know you want to research for example "Health Care", I have it working...Pretty cool.... But I'd like it to do the same for Engineering and I don't know how to change the "topic" or load the topic variable.

@vincentreynolds853 2 ай бұрын

Help??! > sudo pip3 install crewai Could not find a version that satisfies the requirement crewai (from versions: ) No matching distribution found for crewai

@vincentreynolds853 2 ай бұрын

4.18.0-305.el8.x86_64

@bhancock_ai 2 ай бұрын

Hey! I'd recommend to u[date to the latest version of crewAI which is currently 0.76.2. Please let me know if you run into the same issue after updating versions!