Forget OpenAI! Flowise Tool Agents Work with ANYONE (ft. Claude 3)

Рет қаралды 8,176

Күн бұрын

Flowise recently added a Tool Agent to the platform.
The Flowise Tool Agent allows you to attach any AI model that supports function calling or tool use to it. This means that you are no longer limitedto using OpenAI for creating AI Agents and all other major models are supported - like Gemini, Claude, Groq and more.
🙏 Support My Channel:
Buy me a coffee ☕ : www.buymeacoffee.com/leonvanzyl
PayPal Donation: www.paypal.com/ncp/payment/EK...
💬 Chat with Like-Minded Individuals on Discord:
/ discord
🧠 I can build your chatbots for you!
www.cognaitiv.ai
🕒 TIMESTAMPS:
00:00 - Intro
01:18 - Create Anthropic API Key
02:23 - Creating the Tool Agent
03:36 - Claude 3 Models Overview
04:17 - Quick look at Image Upload / Vision
04:55 - Adding Tools
07:51 - RAG and Document Loaders
10:26 - Voyage AI Embeddings
12:55 - Dimensions for Vector Index
14:02 - Outro

Пікірлер: 83

@stealthesethoughts 2 ай бұрын

Great as always. I'd love to see this for Llama 3 and Mistral. I want to leverage more open source models.

@Alvaro-cs7zs 2 ай бұрын

I would love to see that with llama3 as well! It would be amazing a tutorial with Groq API o even in local

@PIOT23 2 ай бұрын

Same

@RoniBliss 10 күн бұрын

@@Alvaro-cs7zsWhy would you need see this? Would it not be better to actually try it for yourself? Its very veryyyy easy guys, its exactly the same process, but instead of here you see where he uses "Claude Chat Model Node"? Well stay with me we going deep and complex.... Instead of the Claude Chat Model node select the Groq Chat model Node, then select the model (Mixtral / Llama 3) from Gtoq node drop down! ... Done 😅!

@nicolasgramnea1324 2 ай бұрын

Thank you, once again!

@lAutoDidacte 2 ай бұрын

Hello and thx for the videos. What about llama 3 ? How to use it with Flowise ? Can we self host the model and integrate it with API to Flowise ?

@traxtibidox5775 2 ай бұрын

Thx for the great video Leon ❤ was waiting for this tutorial 😂

@leonvanzyl 2 ай бұрын

You're welcome 🤗

@estevegraells 2 ай бұрын

Gorgeous, super useful!!

@leonvanzyl 2 ай бұрын

You're welcome 🤗

@sdcharly 2 ай бұрын

Good one as usual..

@WayneBruton 2 ай бұрын

Hi Leon, tried the same as above with Groq instead of Anthropic and Mixtral... It works like a charm. Thanks again

@leonvanzyl 2 ай бұрын

That's an awesome model to use as well 👍

@WilsonEugenio 2 ай бұрын

Very good video! Cheers, from Brazil

@leonvanzyl 2 ай бұрын

Thank you! Howzit from South Africa 🦁

@Luxcium 2 ай бұрын

I didn’t use the Anthropic model extensively but I did find that Haiku was very potent and I personally didn’t find any limitations also I did not make a full comparison with questions and all…

@leonvanzyl 2 ай бұрын

Agreed. Haiku is incredible.

@semahselmi Ай бұрын

i 4ll be forever grateful for you , There's only Value Out here in this Channel!

@YitzhakErezKarat 28 күн бұрын

Hi In the Voyage AI Embeddings how can I choose or add different models like voyage-law-2 ? Thanks in advance

@tecnopadre 2 ай бұрын

Groq. Or Cohere that AFAI it was giving some problems. Also with Cohere re-ranking would be great. What I mean is that with the hard work you are doing, it would be great to have the perfect RAG stack build. And push Flowise to implement neo4j or similiar Graph databases would be awesome. Congrats Leon. 🦁

@CptTurk81 24 күн бұрын

Has anyone got a free solution for using a function calling compatible chat as im getting errors using chatollama function within my agent flow.

@rachdeg1087 2 ай бұрын

Flowise is absolutely great, i'm just missing a proper integration with LM studio

@matteominellono 2 ай бұрын

Knowing the speed of llama 3 on LM Studio (which is very similar to chat RTX) and the fact that LM can create a local server, more agents software should aim towards the direction of adding local LLMs

@rachdeg1087 2 ай бұрын

@@matteominellono exactly this! I agree!

@RiwutLibinuko Ай бұрын

Is that possible to use Ollama for function calling?

@Superdooperhero 2 ай бұрын

Does this work with any local model where you don't need an API key?

@leonvanzyl 2 ай бұрын

If the model supports function calling and you have the hardware, then yes ☺️

@Superdooperhero 2 ай бұрын

@@leonvanzyl And I assume there isn't currently such a free model

@GilbertMizrahi 2 ай бұрын

Great as always. Let's suppose you have a custom tool that transcribes an audio file. The resulting text is larger than 200k that Claude 3 can handle. Is there a way to shunk, embed, and vectorize that text, so a user can ask questions about it?

@maniecronje 2 ай бұрын

You asked for suggestions around AI models would love to see how to use AWS Bedrock, thanks Leon

@leonvanzyl 2 ай бұрын

Excellent suggestion

@gatlashirisha9344 Ай бұрын

Thank you so much @Leon can you pls cover calling image api and document flow together

@jonzh4720 2 ай бұрын

Leon, can we implement dual retrieval systems-one from documents and another from SQL or structured table data-each utilizing distinct embedding models, with the chat automatically routing requests accordingly?

@xavierf2229 23 күн бұрын

Do you think its possible to build a bot like this just using hugging face models ? Thanks for the tips

@lingzhang2061 Ай бұрын

Thank you for this great tutorial. I learned a lot from watching all your videos. I was wondering if you would have another tutorial about creating multi-agent systems using flowise?

@leonvanzyl Ай бұрын

Working on that video as we speak 😊

@lingzhang2061 Ай бұрын

@@leonvanzyl Thank you so much! I literally can't wait to watch the video

@ganeshnikkam9410 2 ай бұрын

Leon, Is it possible to simulate langgraph flow using existing Flowise AI nodes ? If so, shall you please create a tutorial on it

@purvislewies3118 27 күн бұрын

blessedbrother

@ricardofernandez2286 Ай бұрын

Hi León, possibilities open!! Thank you!!!! The dimensions value is mandatory for all vector stores? Everytime I See Pinecone I See reference to dimensiones, but for example I use chromadb and never was asked nor entered a dimensión value. Is this something Flowise takes care of, or it is just that chromadb does not need this value? Thank you!!!

@leonvanzyl Ай бұрын

That's actually a really good question. I don't know 😁. I noticed that the new Upstash Vector also does not require the dimensions, yet it works with all embedding models. Maybe someone in the comments know..

@iyanpixel 2 ай бұрын

Thanks so much for this helpful content, i have a question, I'm using supabase as my database, if you'll like, can you teach us on how to use it as the vector store

@MrTea2097 2 ай бұрын

Really helpful thanks. One question, from my testing it seems you don't get streaming with this setup (there's a long delay then the answer arrives all at once) is that correct?

@leonvanzyl 2 ай бұрын

Well spotted. I think this would depend on the model that you're using. Not all models support streaming.

@yolopookie 2 ай бұрын

What's the best approach to handle 500K+ rows of sales data in CSV format to be able to answer various data analysis questions? I watch all your videos, and all examples are about small Word documents

@innovationscode9909 2 ай бұрын

There is a csv file tool where you can load data from csv.

@WilsonEugenio 2 ай бұрын

Hey friend, have you ever thought about making videos integrating chatbots with AI and N8N?

@leonvanzyl 2 ай бұрын

I have actually. Working on it 😜

@traderbully 2 ай бұрын

I’m just now discovering AI agents and to be honest AI in general. I have never coded before however I’ve had several brainchild applications built for myself and my business. I worked very closely with the developers prior to even starting the project. I left a very little to the imagination and determining, the entire workflow from section the section of the application so I feel that almost this is kind of doing the same thing in a graphical manner. I suppose my question is how difficult is it for someone like me to pick up this type of interface I was able to follow along pretty well on this video I suppose I would just need to know where to find references to all these tools and their functionalities. Would you all consider this a pretty basic video video? Does it get far more complicated than this?

@WilsonEugenio 2 ай бұрын

Leon, is Anthropic cheaper than openai? I didn't understand the Haiku model charges Input: $0.25 / MTok and Output: $1.25 / MTok and openai charges $0.50 / 1M tokens $1.50 / 1M tokens (3.5 turbo) but Anthropic is only 200 thousand tokens while openai is 1M tokens, right ? 4 times more than Anthropic. That's right? I'm looking for a cheaper alternative than openai.

@leonvanzyl 2 ай бұрын

Claude has a 200,000 token context window (can go up to 1 million). I believe it's cheaper as well.

@quentin4488 Ай бұрын

Hello, Thank you for your video. I would like to know if it is possible, instead of having the chatbot on an icon on my site, to have it on a page of my website. Could you tell me how can I do this? Thank you.

@leonvanzyl Ай бұрын

You can use the full page embed for this.

@ednan.martins 26 күн бұрын

Friend, I'm trying to use the retriever tool, but unfortunately the Tool Agent can't "see" it. It completely ignores the Retriver Tool, even though it is completely configured. And what could it be? And what to do to agree? Thank you very much and congratulations for the content

@leonvanzyl 26 күн бұрын

You could try to adjust the system prompt to force it to always use the tool

@ednan.martins 26 күн бұрын

@@leonvanzyl I got it, thank you very much. But I confess that it didn't turn out as well as I wanted, can I find something similar to real estate assistance with flowise on your channel?

@frankbazuaye4747 3 күн бұрын

Hello Leon . I have built a RAG application with flowise .I want to be able to upload a pdf or document and query it , just like chatgpt , or Germini they have the document upload button.Please how can i do that ?

@leonvanzyl 2 күн бұрын

Within Flowise, you will have to use a document loader. It's not possible at this time to upload docs on the chat window. If you're using the Flowise API, then you can load documents as well. Check out my video on using the Flowise API.

@frankbazuaye4747 2 күн бұрын

@@leonvanzyl Thanks Leon I shall watch the video now

@rickyS-D76 2 ай бұрын

Great ❤, please upload Rag chatbot video using weaviate, llama3, where doing semantic search on database column of csv files/real database like supabase or others

@sadyaz64 2 ай бұрын

thanks llam3 Everybody talks about it

@luciameyer2288 Ай бұрын

Hi! I followed the steps along with Leon, and when I try to test the flow with the calculator, I get an API key error: "Your credit balance is too low to access the Claude API. Please go to Plans & Billing to upgrade or purchase credits." I thought it was free. Has this happened to anyone else?

@leonvanzyl Ай бұрын

It is not free. You get an initial $5 of free credit when you verify your account.

@luciameyer2288 Ай бұрын

@@leonvanzyl Oh, I get it. Thanks!

@WayneBruton 2 ай бұрын

Hi Leon, under tools there is requests post... any chance of a tutorial on that?

@leonvanzyl 2 ай бұрын

Will do

@WayneBruton 2 ай бұрын

@@leonvanzyl I actually achieved what I wanted using the custom tool but I would guess the requests post would be easier?

@milesbarn 2 ай бұрын

Can someone explain the reason for embedding vector stores? It just seems like a step that could be avoided altogether but isn’t. Why? Why are all these vector stores popping up to accept our data and documents? And is there a way to solve this with a local vector store? What an epic tutorial, thank you Leon

@leonvanzyl 2 ай бұрын

I actually have a few videos that explain the reason for using Vector Stores. kzfaq.info/get/bejne/jJ2lddxku7eucZc.html In a nutshell, you shouldn't pass the entire knowledge base into the chat as context. Instead, you need to chunk the data and store the chunks somewhere (vector database). Only the most relevant chunks will be added to the context.

@PIOT23 2 ай бұрын

I see new video I like! Always the best content

@leonvanzyl 2 ай бұрын

Haha, thank you 😊

@travistang9677 2 ай бұрын

Can I use local functional model?

@leonvanzyl 2 ай бұрын

If it supports function calling, then yes :-) . Not all models support function / Tool Use yet.

@travistang9677 2 ай бұрын

@@leonvanzyl Could you give me some example models can run locally?

@maniecronje 2 ай бұрын

Thanks Leon great content as always, I am missing the Tool Agent from the menu I assume I need to update Flowise, did you have a video explaining how to do the latest update of Flowise.I followed the instructions downloaded the script you have supplied on GitHub ( kzfaq.info/get/bejne/pNdxfsJ9m5zXeYU.html ) I am not sure why but I can not see the latest Tool Agent in my toolbar, my current version is 1.6.2 I noticed that the latest is 1.6.6 can you please explain how I may upgrade to the latest 🙏🙏? thx Manie

@leonvanzyl 2 ай бұрын

Yes, you need to update Flowise first 👍. My Getting Started video shows how to update locally and the Deploying video shows how to update on Render.

@maniecronje 2 ай бұрын

@@leonvanzyl Baie dankie 🙏🙏

@mike-sf6gl Ай бұрын

Thank You For Your Help. I have a question that Can I Create mutiple types of chatbots like customer support, ecommerce, healthcare, recruitment & many more. Can I integrate flowise chabots in multiple platforms like whatsapp, Instagram and many more. Please answer my question.