VectorShift: Build an AI Agent that Adapts to User Intent (Multi-Step Actions!)

Рет қаралды 3,827

Күн бұрын

In this video we will use VectorShift to build an AI agent that can identify the user's intent and then execute specific tasks for that intent.
This agent will answer questions from a CSV file (ie. a CSV Agent) and use RAG to answer questions from a knowledge base (a PDF in this instance).
🙏 Support My Channel:
Buy me a coffee ☕ : www.buymeacoffee.com/leonvanzyl
PayPal Donation: www.paypal.com/ncp/payment/EK...
📑 Useful Links:
VectorShift: www.vectorshift.ai
💬 Chat with Like-Minded Individuals on Discord:
/ discord
🧠 I can build your chatbots for you!
www.cognaitiv.ai
🕒 TIMESTAMPS:
00:00 - Intro
02:32 - Creating the Pipeline
03:29 - Conditions Node
05:01 - Merge Node
06:11 - Using LLM to Identify Intent
08:20 - Pricing Path
12:16 - "Ford" Path
15:08 - Deploying Chatbot
16:08 - Outro

Пікірлер: 40

@HemaSherbeny Ай бұрын

I cannot pass any video for you Leo, Please dont stop... I see a great youtuber with 1M subscriber in the AI field in 2025

@leonvanzyl Ай бұрын

Thank you!!

@humphuk Ай бұрын

Leon, I think you have elevated to my No1 favourite KZfaqr. I enjoy catching up with the latest news on GAI on other channels - but you help me to use and explore. I find you videos really easy to follow and replicate - a testament to the work you put into planning and editing. While I get a great sense of satisfaction from copying and modifying your examples - I am always left with ...... "well, if it can do that ...... what else could it do? Thinking hat on!

@leonvanzyl Ай бұрын

Awesome comment and feedback! Thank you very much.

@ILoveProjects Ай бұрын

@leonvanzyl, Interesting to see the comparison between Flowise and VectorShift. Any plans? Personally, after your videos I just started using Flowise, and now I see your videos of VectorShift. Which tool to use?

@leonvanzyl Ай бұрын

A comparison video would be a good idea. Can add n8n and Langflow in the mix as well. FW and VS each offer unique features.

@leechoonhwee Ай бұрын

Hi Leon, thanks for the great video. I would like to ask what happens if the user ask what is the price of a ford focus without the year? Will the agent prompt further for the year before continuing the path?

@Alvaro-cs7zs Ай бұрын

Great video Leon! Awesome to see the way conditions work. But to be honest, 25$ for just 5 knowledge bases... I find it too expensive! But as said before, it is a luxury to see your videos Leon. Simple to understand, great value in the content and the way you explain things is exceptional. You have become one of my favourites to watch, it not the most!

@leonvanzyl Ай бұрын

Thank you! Yeah, most of these tools - including the open source one's - come at a cost when you start looking at hosting these platforms. You also don't have to bring your own OpenAI key with VectorShift, which I appreciate as well. My mission is to introduce you to real world applications that you can use for your or your clients projects

@christiansard5247 Ай бұрын

Thanks leon for your videos ;-) all those automation & chatbot platforms are great and you render their usage so easy ;-) but i need those tools fully self hosted, my clients don't want to hear about a cloud solution... will test tiledesk for the chatbot part, n8n for the automation and obviously our beloved flowise for some backend workflows, the inference and models will be llmstudio or Olama and local vector database will be qdrant; do you already use or plan to test some self hosted platforms ?

@leonvanzyl Ай бұрын

Absolutely. I've been meaning to cover n8n for a while now.

@Andersomatheusful Ай бұрын

I love your videos! from Brasil 🇧🇷

@leonvanzyl Ай бұрын

Hello from South Africa 🇿🇦

Ай бұрын

Is the youtube membership active? Curious what member only videos would look like.

@leonvanzyl Ай бұрын

Members get early access to certain videos. I will be creating exclusive content soon. Another idea was to make the chatflows / downloadable content available to members only

@tecnopadre Ай бұрын

On this example you use two different prompts. One general and one for the CSV filtering. Didn't you think about merging on a single prompt? Wouldn't it be cheaper or faster having only one API call?

@leonvanzyl Ай бұрын

For sure, feel free to reuse nodes in your flow. I think the point was that you could create branching paths with different prompt, models and actions 👍

@PubgSpeed-wl8yo Ай бұрын

Tell me, are VectorShift, Buildship, Flowise essentially the same thing, or are some services not included in this three?

@fin101-z3o Ай бұрын

great video. what happens when the user's question has both or multiple intents? which path then?

@leonvanzyl Ай бұрын

You'll have to cater for that intent. It's the same way platforms like Voiceflow and Botpress deals with it. That, or the agent will decide which one of the two paths to take.

@Create-The-Imaginable Ай бұрын

If the AI is smart enough couldn't you have a both paths as part of the logic if they ask for both and provide both answers one after another? That is what a human would do! I think the current LLMs can do this but we should test and try it! Maybe there is a 4th pieces of logic "I do not know, could you rephrase your question!"

@tecnopadre Ай бұрын

Are you seeing Vector Shift more powerful than Flowise and if not, would you elaborate please? I've invest a lot in Flowise, and would like to know pros and cons. If you could. For example, in this video there is no chunking variables, there is no way to manage the PDF parse, etc. Something that sometimes (not always) on Flowise you can. Flowise also can be installed in local. Something that Vector shift as far as I know you can't. I'm seeing Flowise closer to code then VS. Do you? Thank you for such great videos Leon. Keep pushing!!

@leonvanzyl Ай бұрын

You might want to check out the previous videos that I made on VS to get an understanding of the differences. For me personally I like the automation capabilities of VS.

@tecnopadre Ай бұрын

@@leonvanzyl I didn't miss any of your videos 🙂

@MsAntonio664 Ай бұрын

possible to make these interactions in flowise?

@roegger Ай бұрын

Hi, what do you suggest if I have several csv files? In practice I might have hundreds of csv files and also many texts from pdfs etc. do you share the workflows with members?

@leonvanzyl Ай бұрын

It might be best to use a Knowledge base in that instance. You can upload as many files as you want then 😊. That's a good idea actually. I'll start sharing the project files with members. Give me some time to figure out the logistics. Awesome idea!

@maniecronje Ай бұрын

Great content as always, thank you for all your effort producing this content, "BAIE DANKIE", just a quick question can you achieve the same capabilities explained here using Flowise?

@leonvanzyl Ай бұрын

You're welcome 🤗. Absolutely, you can build agents in FW (I have plenty of FW Agent videos on my channel). Things are like RAG are easier to implement and manage using VS though.

@maniecronje Ай бұрын

@@leonvanzyl thank you Leon

@reinhardrinaldo350 Ай бұрын

Apparently the CSV query loader works only with a specific csv format.

@leonvanzyl Ай бұрын

I think you just need to adjust the prompt of the LLM to match the CSV structure

@suhnyllakler5842 Ай бұрын

Leon thanks. Brilliant video. Quick question if possible please - for the CSV - you loaded it via "Data Loaders" and CSV. For the 10K - you used "Knowledge Base" then uploaded the file. But file can also be loaded via "Data Loaders" - when or why did you choose which?

@leonvanzyl Ай бұрын

Vectors are great when dealing with massive amounts of data that need to be chunked. The CSV loader reads the data on the fly so if the file is small enough you don't have to load it into a vector store. In this example I could have just used either solutions, but the main takeaway really is the ability to take the user down different paths.

@suhnyllakler5842 Ай бұрын

@@leonvanzyl ok gotcha. Kind thank you😊

@amrpyt Ай бұрын

is it better than flowis in doing this task?

@marvinmarkham8405 Ай бұрын

Seems to be yes

@PIOT23 Ай бұрын

LVZ as always you’re the lord master of AI content how would you describe the difference in using flowise vs vectorshift Vectorshift seems like it could be easier to get more granular with your flow, would you agree? Or am I talking out my 🍑?

@ILoveProjects Ай бұрын

Agree. In a higly regulated industry (banking in US) the in-house data is crucial. How VectorShift differs from Flowise?

@fiveofknives Ай бұрын

Hi Leon, I have sent 3 emails via your contact form but no response, would love to work with you do you perhaps have linked in or a place where i can contact you? X - cannot DM, Contact form on your website either does not work or emails being ignored - Cheers