Thanks for the video Leo, but if you can do more example of Sequential, we would like it, for example like a conversation with a customer like for sales, but they want to purchase a product, so that the bot has the intention of seeing when it has to ask them. the client's data, of course I see that it is complex, but the most possible.
Leon van Zyl, thank you for your valuable videos, very informative. I learned alot from your content. I wanted to ask, are you able to make a video that shows how to use N8N with Flowise and host them both in own cloud? thank you,
I just deployed this last night in AWS. Quick note, make sure you use more than a 8gb disk (I ran out of space, worked when I used 28gb) and also use the Docker instructions (NPM froze on me installing both for some reason).
@@leonvanzyl I have some workflows written in langgraph, which is better no code framework to use instead of langgraph ? Or can I use langgraph via api with n8n ? Thank you
I would love to know how to send additional parameters (like "name") from Telegram to Flowise. I can currently only get sessionID to work (for Upstart Redis).
Great Video Leon, a little more work with Sequential agents vs multi agents but I think the effort is well worth it unless you know the demands won't grow. I have been very keen on using Assistant API but this gives more flexibility. For a simple RAG Bot, its pretty hard to beat the Assistants I guess, quick output, quick money (esp with 40 Mini). But the sequential agents is so very powerful. Thank you.
The Sequential agents are amazing! More videos about It please. PD: Also how to use Flowise and Langchain at the same time and the benefits of using both together ;) Thanks, Leooo!!
spot on, you are my go to for understanding flowise! Any chance you can do a tutorial on how to do an integrated code dev application using flowise...is this even possible? I want to achieve something like open interpreter, but with all the flowise no-code / flexible flow. Maybe you could build on your earlier multi agent sw dev tutorial to actually debug and review the code?
In Python it is possible chat with Excel files using langchain and pandasai (storing the excel data into a SmartDataframe). Do you know if it is possible do something similar with flowise?
Awesome, Leon! Thanks so much! This is incredibly useful as always. I have a special use case I'd love to explore with sequential agents. I'm thinking of one agent tasked with collecting information from users (e.g., filling out a form with a defined number of variables) and another agent responsible for processing the information collected to provide any type of conclusion to the user :)
@jhonymonhol6851 I'm cleaning up some sensitive data out of it,, and I will share as a gist today. I'm actually getting VERY CLOSE to making it work quite nice (but needs work), I'm sure we can collab and make it perfect!
I think sequential is so complicated seeing that the previous squema does the same.. don't you think? What would be the advantages or disadvantages of both?
What is this error about? Error buildAgentGraph - Cannot read properties of undefined (reading 'label') I set two agents with a retrieval tool connected to pinecone for each one. Followed the same steps as the video.
Followed the video step-by-step and got that same error message Error buildAgentGraph - Cannot read properties of undefined (reading 'label') Wondering how you can debug Flowise when errors like this come up?
I'm actually going to create a video on this. It's possible, but I think it requires a bit of explanation. Nothing too extreme though. Involves some state variables, LLM nodes, memory and conditions.
@@leonvanzyl thank you so much! This is a really important architecture for many tasks, since the LLMs typically need human guidance and confirmation to promote better results.
Hey Leon... Why in both examples do you instructs the worker to pass the result to the other worker (12:25), but it always goes back to the supervisor (who takes care of it)!?
I'm using an OpenAI model in this video, so my specs don't really matter. If speed is impoyto you then I highly recommend giving Groq a try. I've got a dedicated video on Groq.
Good one again @Leon.. Can i ask how LLM node and Condition Agent is different based on this video?.. might be a lame query :) or shall i put it like is LLM Node + Condition = Condition agent?
Thanks for the video, Leon! I was wondering if it is possible with Sequential Agents to set a condition that "one agent finishes its task"? I want one agent to ask the user about 10-15 questions to gather a lot of information and then send it to another agent that will analyze the information before responding to the user. Do I need to create 10-15 nodes to store each piece of information? Thanks if you have time to answer, and thank you very much for all you videos! It helps a lot !!
Thank you for the feedback! You could probably create your example using a normal agent in Flowise. I guess it depends on what the second agent in the chain needs to do. I would honestly suggest that you first try to create your example using a normal flow with a tool agent. Agentflows might not be a good match for collecting information from a user in a loop.
I really love your videos. This is no exception. Do you have a Discord or something? I am really interested in getting sequential agent working with Llama 3.1 but having issues. Again, awesome video and thanks.
Excellent tutorial! 👏 I've been trying to get Flowise API to work in an app much like a pipeline outside the chatbot. According to the documentation you can only stream output with a Chain or Tool Agent. I attempted to use a Sequential Agent like in your example but it only returns the final result and not the stream itself. Do you think you can do a video on how to correctly accomplish this? Thanks!
Thank you Leon. Great video. I can see so much potential for this. I did run into an issue however, duplicating your sequential workflow (exact, except for using ChatOllama, instead of ChatOpenAI). When I run the flow, the software developer is called, as it should, stating that the code needs to be reviewed, but then i get the error: Error buildAgentGraph - Error: No tool calls found in the response. and the flow just stops there. Any suggestions?
@@leonvanzyl i just changed to GrogChat (llama 3.1 70b) and that seems to work better, although the software developer seems to think it has done a good enough job and does not need to do more, after the code review, even if the code reviewer suggested more changes :D
Hi Leon apologies if I already asked this in earlier videos… can FW handle calling JS function in the same way as shown here czcams.com/video/HsWwyipZNHw/video.htmlsi=i8fHxMZmliGzBcwz
Absolutely. The custom tools in FW run in a node environment, so you can add JavaScript code. I created a video on this. It's one of the older videos, but tools still work the same way. czcams.com/video/_K9xJqEgnrU/video.html
@@leonvanzyl Thx, Leon is there a way to use GraphRAG with FW? if it is possible could you create tutorial on this or point me what you think would be good starting point, thx Manie
Until Flowise add streaming capabilities to AgentFlows and Sequential agents they are pretty much redundant. As through API the responses are delivered in one huge block of text. Making it unusable for production applications.
I kinda agree with you to a smaller extent. Consider this: You could call the API from a frontend (like a chatbot) or via an automation / non client facing process. If you're using it as a chatbot via the web embed, then I agree. It would be a bit frustrating to wait for a response. Flowise will definitely add streaming for this, so it won't be an issue for long. If you're using it during a workflow or backend process then streaming really doesn't matter. For example, you could have some research team execute at 9am every morning using Make, Zapier or n8n. The final result will then be emailed to you or stored in a database. I suspect that this will be the most common use-case for Agentflows.
Thanks for the video Leo, but if you can do more example of Sequential, we would like it, for example like a conversation with a customer like for sales, but they want to purchase a product, so that the bot has the intention of seeing when it has to ask them. the client's data, of course I see that it is complex, but the most possible.
I've just started with the beginner's guide, so this is a bit more complex, but I'm amazed at the possibilities 😧
Thank you for your high value content ❤💯🙏 looking forward for the n8n tutoriels too 🎉
You're welcome 🤗.
Leon van Zyl, thank you for your valuable videos, very informative. I learned alot from your content. I wanted to ask, are you able to make a video that shows how to use N8N with Flowise and host them both in own cloud? thank you,
That's actually my favourite stack! I have both Flowise and n8n running on Render with some automation.
I'll create a video on this.
I just deployed this last night in AWS. Quick note, make sure you use more than a 8gb disk (I ran out of space, worked when I used 28gb) and also use the Docker instructions (NPM froze on me installing both for some reason).
@@leonvanzyl very much looking forward to this.
@@leonvanzyl I have some workflows written in langgraph, which is better no code framework to use instead of langgraph ? Or can I use langgraph via api with n8n ? Thank you
I would love to know how to send additional parameters (like "name") from Telegram to Flowise. I can currently only get sessionID to work (for Upstart Redis).
You can try adding variables inside of the "question" parameter when querying Flowise.
awesome, another awesome video! looking forward to improved rag examples
Coming SOON 😁
Great Video Leon, a little more work with Sequential agents vs multi agents but I think the effort is well worth it unless you know the demands won't grow. I have been very keen on using Assistant API but this gives more flexibility. For a simple RAG Bot, its pretty hard to beat the Assistants I guess, quick output, quick money (esp with 40 Mini). But the sequential agents is so very powerful. Thank you.
Really nice video once again Leon 😊 I was waiting for the final app running 😂 how did it go?
Thank you for the feedback 🙏.
Hahaha, I'm sure the code would have worked 😂
The Sequential agents are amazing! More videos about It please.
PD: Also how to use Flowise and Langchain at the same time and the benefits of using both together ;)
Thanks, Leooo!!
You are a very awesome mentor! thank you! 🎉
spot on, you are my go to for understanding flowise! Any chance you can do a tutorial on how to do an integrated code dev application using flowise...is this even possible? I want to achieve something like open interpreter, but with all the flowise no-code / flexible flow. Maybe you could build on your earlier multi agent sw dev tutorial to actually debug and review the code?
Thank you for the suggestion.
I'll look into this 👍.
Thank you very much for the super! It helps a lot 🙏
I would love if you showcase the code, to see if the results! And if the app works. Amazing content really 🙏
Noted.
@@leonvanzyl I meant to test the code in the video! Maybe I expressed wrongly
Thanks Leon! Great tutorial!
Thank you 🙏
Great video! I see you use both Flowise and Langflow. Which one is your favorite? For production use
In Python it is possible chat with Excel files using langchain and pandasai (storing the excel data into a SmartDataframe). Do you know if it is possible do something similar with flowise?
Awesome, Leon! Thanks so much! This is incredibly useful as always.
I have a special use case I'd love to explore with sequential agents. I'm thinking of one agent tasked with collecting information from users (e.g., filling out a form with a defined number of variables) and another agent responsible for processing the information collected to provide any type of conclusion to the user :)
@choistella5863, I am working on exactly the same, I could share my current flow, maybe it would help.
@@LuisOrtizFX
Thanks so much for your answer !!!
It would be amazing if you could share your flow :)
@@LuisOrtizFX I would like to take a look, how can you share this with us?
@jhonymonhol6851 I'm cleaning up some sensitive data out of it,, and I will share as a gist today. I'm actually getting VERY CLOSE to making it work quite nice (but needs work), I'm sure we can collab and make it perfect!
Thanks for the video, one question: Where can I define the system context, for receive the user? Thank you so much
I think sequential is so complicated seeing that the previous squema does the same.. don't you think? What would be the advantages or disadvantages of both?
Great video as always.
Thank you
Maaaan you da MvP! Thanks
What is this error about?
Error buildAgentGraph - Cannot read properties of undefined (reading 'label')
I set two agents with a retrieval tool connected to pinecone for each one. Followed the same steps as the video.
Followed the video step-by-step and got that same error message
Error buildAgentGraph - Cannot read properties of undefined (reading 'label')
Wondering how you can debug Flowise when errors like this come up?
Very cool Leon. Do you have website where you share this flows? Or maybe we can buy some from you :-)
Thank you. I'll see if I can share them on a GitHub repo or somewhere.
Is it possible to loop back to the human for confirmation or feedback? I.e. human in the loop 😊
I'm actually going to create a video on this. It's possible, but I think it requires a bit of explanation. Nothing too extreme though. Involves some state variables, LLM nodes, memory and conditions.
@@leonvanzyl thank you so much! This is a really important architecture for many tasks, since the LLMs typically need human guidance and confirmation to promote better results.
Have you ever tried with Azure open AI? It works similar?
Hey Leon... Why in both examples do you instructs the worker to pass the result to the other worker (12:25), but it always goes back to the supervisor (who takes care of it)!?
I have found that the supervisor would sometimes end the flow early if you don't provide clear instructions.
@@leonvanzyl Thank your!
Hey Leon, always incredible content. Are you avaiable for hire?
Thank you. You can contact me via email (in about section) or my automation agency (in video description).
@@leonvanzyl Great, sent you an email.
What are the system specs are you running these models on? It looks super snappy when they respond
I'm using an OpenAI model in this video, so my specs don't really matter.
If speed is impoyto you then I highly recommend giving Groq a try. I've got a dedicated video on Groq.
Good one again @Leon.. Can i ask how LLM node and Condition Agent is different based on this video?.. might be a lame query :) or shall i put it like is LLM Node + Condition = Condition agent?
Actually, that's a very good way to look at it.
The LLM Node can return a structured output which a condition node can then use.
Thanks for the video, Leon! I was wondering if it is possible with Sequential Agents to set a condition that "one agent finishes its task"?
I want one agent to ask the user about 10-15 questions to gather a lot of information and then send it to another agent that will analyze the information before responding to the user. Do I need to create 10-15 nodes to store each piece of information? Thanks if you have time to answer, and thank you very much for all you videos! It helps a lot !!
Thank you for the feedback!
You could probably create your example using a normal agent in Flowise. I guess it depends on what the second agent in the chain needs to do.
I would honestly suggest that you first try to create your example using a normal flow with a tool agent.
Agentflows might not be a good match for collecting information from a user in a loop.
I really love your videos. This is no exception. Do you have a Discord or something? I am really interested in getting sequential agent working with Llama 3.1 but having issues. Again, awesome video and thanks.
Thank you! I actually suggest joining the Flowise AI dievord directly.
Excellent tutorial! 👏
I've been trying to get Flowise API to work in an app much like a pipeline outside the chatbot. According to the documentation you can only stream output with a Chain or Tool Agent. I attempted to use a Sequential Agent like in your example but it only returns the final result and not the stream itself. Do you think you can do a video on how to correctly accomplish this? Thanks!
I think the agentflows cannot be streamed at this stage.
@leonvanzyl I see. I even tried it with a Chain too but did not work as well.
Thank you Leon. Great video. I can see so much potential for this. I did run into an issue however, duplicating your sequential workflow (exact, except for using ChatOllama, instead of ChatOpenAI). When I run the flow, the software developer is called, as it should, stating that the code needs to be reviewed, but then i get the error: Error buildAgentGraph - Error: No tool calls found in the response. and the flow just stops there. Any suggestions?
You need to use an advanced model for multi-agents. If you're using Ollama, it will have to be the equivalent of the Llama 3.1 70b model.
@@leonvanzyl Ah great. Thank you. Will give that a go then (or use a non-local one).
@@leonvanzyl i just changed to GrogChat (llama 3.1 70b) and that seems to work better, although the software developer seems to think it has done a good enough job and does not need to do more, after the code review, even if the code reviewer suggested more changes :D
Hey how do I join your channel?….can’t see any options to be a member? Thx
You should see a "Join" button somewhere on any of my vids.
@@leonvanzyl hey weirdly it’s not showing up on iPad…fine on android…FYI…
Am I crazy or is some of this out of date already? I can't find the start, LLM, or the state node... whats up?
What version of FW are you on? These nodes were added with v2
@@leonvanzyl If you get someone else who asks this... I was in Chatflow NOT Agentflow. User error in the biggest way.
Hi Leon apologies if I already asked this in earlier videos… can FW handle calling JS function in the same way as shown here czcams.com/video/HsWwyipZNHw/video.htmlsi=i8fHxMZmliGzBcwz
Absolutely.
The custom tools in FW run in a node environment, so you can add JavaScript code.
I created a video on this. It's one of the older videos, but tools still work the same way.
czcams.com/video/_K9xJqEgnrU/video.html
@@leonvanzyl Bakgat! Baie Dankie!
@@leonvanzyl Thx, Leon is there a way to use GraphRAG with FW? if it is possible could you create tutorial on this or point me what you think would be good starting point, thx Manie
Until Flowise add streaming capabilities to AgentFlows and Sequential agents they are pretty much redundant. As through API the responses are delivered in one huge block of text. Making it unusable for production applications.
I kinda agree with you to a smaller extent.
Consider this:
You could call the API from a frontend (like a chatbot) or via an automation / non client facing process.
If you're using it as a chatbot via the web embed, then I agree. It would be a bit frustrating to wait for a response. Flowise will definitely add streaming for this, so it won't be an issue for long.
If you're using it during a workflow or backend process then streaming really doesn't matter.
For example, you could have some research team execute at 9am every morning using Make, Zapier or n8n. The final result will then be emailed to you or stored in a database. I suspect that this will be the most common use-case for Agentflows.