Thanks for one more great video. However, I would like you guys to show how to work with RAG and Memory with crew AI, that would really be a step up to pro, there are not many docs available that get valuable insights on those two topics. I am considering other frameworks in which connecting an gent to memory is a breeze
@@bhancock_aiIf you could also explore the knowledge option to give crews more context, it would be great, as it is becoming a trend to make the crews learn as they go
Thanks for this. Very clear explaination. How do we determine what kind of instructions belongs in the Description vs the Expected_Output? Are there guidelines to follow?
Are there other ways to pass the inputs to agents executing the tasks rather than injecting them into the description ? IMO ideal way should be, task descriptions should be a free text and there must be a way to pass this inputs object to crew so agents can use them in tasks when required
If I had an AI setup that would help guide and code the CrewAI, that would be worth something. Dont mind designing and guiding, but dont really want to invest time learning unique coding aspects to use CrewAI - that is so 2023
I would definitely recommend checking out our new CLI tool so that you only have to type out `crewai create crew ` to spin up a new crew. From there, you don't even have to deal with code. You primarily just have to update the agent and task yaml files with instructions. Also, there is a free crewAI GPT that helps you create crews! I'll add a link to it down below! docs.crewai.com/quickstart chatgpt.com/g/g-qqTuUWsBY-crewai-assistant
I have a similar series on this- this is a good start, so you are using the new flow for crewAI, but when you create the template with poetry are you discovering that there are lots of linkage areas mixed up in the project?
What if it doesn't look like a task can be divided further? What if it's supposed to be a long task with lots of edge cases or nuances that need to be mentioned?
This is an awesome question! My first suggestion would be to pass over the task over to o1 or some of the really smart models and see if they can break apart the task in to smaller linear subtasks. If there are a bunch of edge cases, that's already a sign you can break up the task into at least 2 parts. - part 1 would be the core activities that need to be performed regardless of the edge cases - part 2 would be "handle edge cases" Also, I have no idea what your crew is doing so not all of this advice might be relevant. Please feel free to shoot your crew over to me if you would like me to potentially analyze it in an upcoming TH-cam video!
Please can someone help me with how to build crews with Gemini models. the tutorial I followed is old it seems that method doesn't work anymore. Any pointers will be appreciated ❤
If you head over to our docs: docs.crewai.com/concepts/llms#google there is a quick code snippet that shows you can setup a gemini LLM in just a few lines of code by updating your .env file and creating a new LLM! Please let me know if you have any other questions!
Why should I make an effort to partition tasks and partition subtasks instead of activating planning? What is the difference between "planning = true" and your pro crew structure?
Thanks for one more great video. However, I would like you guys to show how to work with RAG and Memory with crew AI, that would really be a step up to pro, there are not many docs available that get valuable insights on those two topics. I am considering other frameworks in which connecting an gent to memory is a breeze
Thanks Oliver! Great suggestion!
I can look into doing a noob vs pro episode on RAG soon!
@@bhancock_aiIf you could also explore the knowledge option to give crews more context, it would be great, as it is becoming a trend to make the crews learn as they go
Don’t forget one that use memory !!! 💪🏻
Echo that. RAG + Memory with crew AI. That will be super cool 🎉
Excellent explanation and example of how to break the tasks down. I love the Crew AI stuff and can't wait to give it a try. Thanks for the content.
Great overview and very detailed explanation. I implemented the discussed ideas in my current workflow. Results are amazing...
Love hearing this feedback! Thanks for letting me know 😁
Thanks for this. Very clear explaination. How do we determine what kind of instructions belongs in the Description vs the Expected_Output? Are there guidelines to follow?
CrewAI or LangGraph? Can you explain the differences?
Are there other ways to pass the inputs to agents executing the tasks rather than injecting them into the description ? IMO ideal way should be, task descriptions should be a free text and there must be a way to pass this inputs object to crew so agents can use them in tasks when required
If I had an AI setup that would help guide and code the CrewAI, that would be worth something. Dont mind designing and guiding, but dont really want to invest time learning unique coding aspects to use CrewAI - that is so 2023
I would definitely recommend checking out our new CLI tool so that you only have to type out `crewai create crew ` to spin up a new crew.
From there, you don't even have to deal with code. You primarily just have to update the agent and task yaml files with instructions.
Also, there is a free crewAI GPT that helps you create crews! I'll add a link to it down below!
docs.crewai.com/quickstart
chatgpt.com/g/g-qqTuUWsBY-crewai-assistant
All your contents are excellent. Congratulations!
I have a similar series on this- this is a good start, so you are using the new flow for crewAI, but when you create the template with poetry are you discovering that there are lots of linkage areas mixed up in the project?
What if it doesn't look like a task can be divided further? What if it's supposed to be a long task with lots of edge cases or nuances that need to be mentioned?
This is an awesome question!
My first suggestion would be to pass over the task over to o1 or some of the really smart models and see if they can break apart the task in to smaller linear subtasks.
If there are a bunch of edge cases, that's already a sign you can break up the task into at least 2 parts.
- part 1 would be the core activities that need to be performed regardless of the edge cases
- part 2 would be "handle edge cases"
Also, I have no idea what your crew is doing so not all of this advice might be relevant. Please feel free to shoot your crew over to me if you would like me to potentially analyze it in an upcoming TH-cam video!
Really well explained, thank you for this, keep em coming!
Thanks Mariano! I already have 2 more in the queue!
Also, people are already sending over their crews so I'm super excited to analyze those as well!
Please can someone help me with how to build crews with Gemini models.
the tutorial I followed is old it seems that method doesn't work anymore.
Any pointers will be appreciated ❤
If you head over to our docs: docs.crewai.com/concepts/llms#google
there is a quick code snippet that shows you can setup a gemini LLM in just a few lines of code by updating your .env file and creating a new LLM!
Please let me know if you have any other questions!
@bhancock_ai
thanks man
is one of your videos that got me really interested in agentic system.
Keep up the good work 🔥💯
Excellent content! I’m looking forward to watching the next videos
Thank you
Why should I make an effort to partition tasks and partition subtasks instead of activating planning? What is the difference between "planning = true" and your pro crew structure?
hello Brandon! I'm from Brazil and I need some help/tips with crewai.
Great video 👍
Thanks Douglas!!
Happy New Year, Brandon! 2025 - Let's GO!
Thanks Greg!! So pumped for 2025! It’s gonna be a great year!
Can you do a video on Kaiban js?
Awesome!
Quando terá dublagem automática em seus vídeos? 🥲🇧🇷🇧🇷🇧🇷🇧🇷
💯👏