Join my SKOOL community here: www.skool.com/the-ai-agent-5174/about It's a community for AI Agents and I can better answer questions. If you would like: 👉 developing these agents 👉 understanding workflows with them 👉 using CrewAI as the framework to accomplish this 👉 be a part of a community of like-minded individuals
Hi Tyler! I thought this was really nice. The flows seem like a super powerful feature. It would be really interesting to see more complex tasks as well. I know what I'm spending my evenings on this next week 🛠
The flow structure is a super feature, I'm sure it will be of much help when developing more sophisticated crew flows. Awesome stuff dude, please keep up the great work!
@TylerReedAI -- Thanks for a great video . the "retry" patters does not seems to work ? have you tested the router decoration with retry pattern ? can you please consider doing a video to show us of iterative improvement cycle ?
The flow feature appears to imitate the state management originally introduced by LangGraph. While CrewAI is known for its straightforward and easy-to-understand approach, this new flow feature adds complexity to the process, making the syntax more convoluted compared to LangGraph. If an AI agent essentially operates as a state machine, what kind of agent framework should we adopt? Should we opt for high-level control, like CrewAI, or more low-level control, as seen in LangGraph?
I agree with that, but at least you are not force to use those flows and control the crews as you like. It's not a breaking change if I understood this properly.
so by using flow can I manage the internal state of my crew? I mean to say This allows me to analyse how my agents communicate with each other and which output is passed from one agent to another agent?
can anyone help me to like use local ollama embedding model for custom memory .. embeddings ... i cant set memory of the crew : crew = Crew( agents=[vectorstore_researcher], tasks=[document_retrieval], process=Process.sequential, cache=cache, verbose=verbose, memory=True, embedder={ "provider": "ollama", "config": { "model": "mxbai-embed-large:latest" } } )
Hey no worries! A flow is a combination of just regular python functions, but inside each function you can just have normal code do something like save to a file, or run a crew....or a combination of both! Yes a Crew is a bunch of agents
![How I Automated News with CrewAI... | [Day 6 of 8]](http://i.ytimg.com/vi/dVohE7HjDWw/mqdefault.jpg)
![How I Automated News with CrewAI... | [Day 6 of 8]](/img/tr.png)
Join my SKOOL community here: www.skool.com/the-ai-agent-5174/about
It's a community for AI Agents and I can better answer questions.
If you would like:
👉 developing these agents
👉 understanding workflows with them
👉 using CrewAI as the framework to accomplish this
👉 be a part of a community of like-minded individuals
Hi Tyler! I thought this was really nice. The flows seem like a super powerful feature. It would be really interesting to see more complex tasks as well. I know what I'm spending my evenings on this next week 🛠
thank you! And yes flow does seem really powerful. I will have more complex tasks in the near future 👐
The flow structure is a super feature, I'm sure it will be of much help when developing more sophisticated crew flows. Awesome stuff dude, please keep up the great work!
Yes it is amazing, I think it will be very helpful as well! Thank you
Thank you for such great content. I always enjoy your videos and practical hands on guides
Thanks for watching I appreciate it 🙏
@TylerReedAI -- Thanks for a great video . the "retry" patters does not seems to work ? have you tested the router decoration with retry pattern ? can you please consider doing a video to show us of iterative improvement cycle ?
Astonishing tutorial!!!
Thank you 🙌
The flow feature appears to imitate the state management originally introduced by LangGraph. While CrewAI is known for its straightforward and easy-to-understand approach, this new flow feature adds complexity to the process, making the syntax more convoluted compared to LangGraph. If an AI agent essentially operates as a state machine, what kind of agent framework should we adopt? Should we opt for high-level control, like CrewAI, or more low-level control, as seen in LangGraph?
I agree with that, but at least you are not force to use those flows and control the crews as you like. It's not a breaking change if I understood this properly.
so by using flow can I manage the internal state of my crew? I mean to say This allows me to analyse how my agents communicate with each other and which output is passed from one agent to another agent?
You prefer CrewAI, Langgraph, or Autogen?
CrewAI
can anyone help me to like use local ollama embedding model for custom memory .. embeddings ... i cant set memory of the crew : crew = Crew(
agents=[vectorstore_researcher],
tasks=[document_retrieval],
process=Process.sequential,
cache=cache,
verbose=verbose,
memory=True,
embedder={
"provider": "ollama",
"config": {
"model": "mxbai-embed-large:latest"
}
}
)
Sorry, my Python level is still low. From my understanding, Flow is a bunch of Crew, and the Crew is a bunch of Agents, right?
Hey no worries! A flow is a combination of just regular python functions, but inside each function you can just have normal code do something like save to a file, or run a crew....or a combination of both! Yes a Crew is a bunch of agents