Build a FREE AI Chatbot with LLAMA 3.2 & FlowiseAI (NO CODE)

[@leonvanzyl]

What do you think about the new Llama 3.2 models? Remember to hit the like button and subscribe to boost the video in the YT algorithm.

[@lingzhang2061]

Awesome video! Thanks, Leon. I was wondering if we could use Llama 3.2 models in the sequential multiagent chatflow?

[@AaronBlox-h2t]

Cool video but Ollama sucks lemons....ON both my intel Macbook Pro and win10 Intel Arc A770 16GB, 64GB DDR4 RAM, 5000TB 7400MB/sec ssd system. But that's cuz I choose Arc, but hey, if you have Nvidia and M1 to M3 it shoudlwork fine. Regarding LLama3.2, they are great, downloading thme from Hugginface with a cusotm python script.

[@errorsinconduct]

thanks for the tutorial, everything else works except the chatflow which says "fetch failed" when I try to prompt it. The document side works and retrieves the 4 most relevant documents. What could be the problem?

[@pavanj2238]

@leonvanzyl fetch failed is the error message in chat window

[@msclt1982]

Thank you so much Loe, You videos are crisp and clear. I have created a chatbot, as shown in the video, using LLaMA and configured it on my VM. The chatbot has memory and is responding to questions as expected. However, the problem I am facing is that the bot takes a lot of time to load and respond to each query. I have already built a chatbot with OpenAI, and while it was a different approach, I didn’t encounter this kind of slowness. Anybody faced this issue? Do you have any solution for this?

[@ygdiget4119]

I really love your videos! I would really love to see an in-depth video of you comparing Langflow, Flowise, n8n, VectorShift, etc!

[@dliedke]

Very nice, I followed the guide and worked just fine. Thanks a lot for the detailed tutorial. I generated some document data samples with Claude :D

[@leonvanzyl]

Great to hear!

[@msclt1982]

I have created a chatbot, as shown in the video, using LLaMA and configured it on my VM. The chatbot has memory and is responding to questions as expected. However, the problem I am facing is that the bot takes a lot of time to load and respond to each query. I have already built a chatbot with OpenAI, and while it was a different approach, I didn’t encounter this kind of slowness. Anybody faced this issue? Do you have any solution for this?

[@rauldiaztorres]

Great Leon !!! Thanks for the tutorial, very simple and work fine ! 🏅🏆🎖

[@leonvanzyl]

You're welcome 🤗

[@researchandbuild1751]

I got it working locally, nice video thank you so much for showing this. I wonder though how could I embed this into a product? Seems that I have to at least run a node server to run the chatbot?

[@brianglong11]

Outstanding! I have set up a knowledge base of data and information collected by forestry company and want to extract relevant information for annual reports. Is there a way to create a flowiseAI with a template of queries that I could run to populate the annual report? Thanks again!

[@wgaitanartunduaga]

Thanks Leo, great tutorial. It does a wonderful job. Please can you make a video where the chatbot has the option to save the conversation history and export it as a PDF or TXT or DOCX file.

[@BrianDevJourney]

Great tutorial Leon!! I reached out to your email about a sponsor opportunity. Cheers!

[@leonvanzyl]

Saw the mail. Thank you. Will respond soon.

[@KiWelten]

Amazing tutorial, exactly what I was looking for!

[@leonvanzyl]

You're welcome 🤗

[@Kafuixx]

Hi Leon,love this video as a coding noob looking to create an ai enterprise app for a business. Is it possible using this method. And if so how can this be embedded into that product

[@YanasChanell]

Thank you for the video. What are the minimal system requirements to run all of this on PC?

[@leonvanzyl]

I think you can run this on a potato. Seriously, we tested the model on a 10 year old Acer Spin with an i3 processor, and onboard graphics, and it works.

[@donutguy1281]

Fantastic tutorial! Exactly what I needed. Chrome's AI generated results referenced you on this topic, lol

[@leonvanzyl]

For real!? Haha, that's awesome 😎

[@maniecronje]

Thank you Leon another amazing FW tutorial -unfortunatelly the hallucination issues using RAG still an issue not something you can use in PROD unfortunatelly - looking fwd to next FW video 'dankie Leon'

[@leonvanzyl]

Glad you liked it! Let me know what you'd like me to cover next 👍

[@johannesdeboeck]

@@leonvanzyl If you allow me to answer that question :) A video covering Flowise + a knowledge graph would be great! This would reduce LLM hallucinations as well

[@LamaChandrasena]

Very useful. Thanks. Is it possible to automate document loading (as new docs become available), for example by running a script or copying files into a folder etc, or it has to be manually done everytime through the UI as you demonstrated? Thanks.

[@SoyPorteroYT]

This would be really useful

[@boteeeh]

Just watch his n8n tutorials, there he made with google docs.

[@AaronBlox-h2t]

Sure...look at the source code and integrate it. Oh wait, it's done via Flowise.....

[@SoyPorteroYT]

@@AaronBlox-h2t limitations of no-code/low-code

Thank you. Any chance doing the video how to deploy this and use it in the cloud or on a website?

[@youssefsaidi5921]

Ty! Can you do a tutorial on how you can host and deploy a lead generation chatbot on a website? Thanks!

Sorry, found it on your channel already!

[@MyFukinBass]

Amazing. Thanks for the content. Is it possible to take this bot elsewhere? Say if I ever wanted to have a bot like this on an app like Slack or Discord?

[@mortu90]

Hi Leon, Thank you so much for all the work you’re doing-it’s greatly appreciated! I’ve just become one of your subscribers and I’m really looking forward to your tutorials. I have a question that’s been driving me a bit crazy: Is it possible to implement a chatflow developed in FlowiseAI to work with the frontend of OpenWebUI? I tried to create a pipeline and uploaded it to OpenWebUI but couldn’t get it to work. Is this achievable? Thanks again!

[@mortu90]

As a follow-up, I was also wondering about integrating a chatflow developed in N8N. I feel that combining the power of N8N or Flowise with OpenWebUI would be incredibly powerful to leverage their awesome UI capabilities.

[@leonvanzyl]

Thanks for the sub 💪. I don't think there is a way to add Flowise to Open Web UI. May I ask why you would want to use Web UI instead of the Flowise/ Chatflow UI?

[@mortu90]

@@leonvanzyl Thanks for getting back to me! I thought the Web UI might offer more flexibility for managing user permissions (but I could be wrong!). My idea is to leverage Flowise’s power to build the chatflow logic and then use WebUI to manage access and provide a more polished frontend experience for users. Does that sound like a good approach, or do you think I’m missing something?

[@AhmedMeklad]

Wow, I think this is an advantage for the developers and the geeks.

[@leonvanzyl]

It's awesome.

[@laflammedenis]

Great video. Standing by for my Flowise invite - would love to see "How to Fine Tune llama 3.2"

[@leonvanzyl]

Flowise invite? It's free, and we used the free version in this video. If you're referring to their paid cloud service then you can skip the waiting list by using my affiliate link.

[@henriquemoniz2091]

Is flowise collecting any info (resources based on metadata and namespaces) that could potentially extract private information?

[@leonvanzyl]

No, you can literally disconnect your internet connection and run this solution fully local.

[@ariclenescorreia6048]

Thank you so much for this video it is just what I needed

[@jakkalsvibes]

Thank you Leon for the detailed explanations, Just need to know is there a bug Clearing Chat History or am I missing something

[@leonvanzyl]

Hey Jakkalsvibes (I like your channel by the way 🔥). I'm not aware of any issues. What's issue are you experiencing?

[@jakkalsvibes]

@@leonvanzyl In die Chatflow Chat Screen as ek Clear Chat druk kry ek 'Error: chatMessagesService.removeAllChatMessages - SQL_READONLY: attempt to write a readonly database'

[@jakkalsvibes]

@@leonvanzyl Dankie ek like jou channel ook 🦊. As ek 'Clear Chat' druk kry ek : Error: 'chatMessagesService.removeAllChatMessages - SQLITE_READONLY: attempt to write a readonly database'

[@jakkalsvibes]

​@@leonvanzylsorted

very good explanations here. thank youuu.. can i ask whats the minimum hardware requirements to run these locally? do we need a tensor chip and cuda gpu, how much vram gpu, and pc ram for this ?

[@leonvanzyl]

This model is tiny (like 2GB). It should run on a potato 😄

@@leonvanzyl 🤣🤣🤣🤣🤣🤣🤣 i adored this answer.. thank youuu 🤘🤘

[@Augmented_AI]

Do you have a graph rag workflow?

[@leonvanzyl]

Not yet. Will create one soon 😄

[@Augmented_AI]

@@leonvanzyl Lekker, looking forward to it :).

[@adrianobernagozzi-ik1tecno362]

Great video, I'm doing a training but the text I'm using is in txt format and is very large, sometimes the questions have answers that make sense and other times they don't. Do you have any tips?

[@JoaoCarlos-bp8ip]

Thanks Leon, another great tutorial. What is the minimum hardware requirements to run this setup?

[@leonvanzyl]

I think you can run this on a potato 🤣. Seriously, I managed to get this running on an 5 year old i3 laptop with an onboard GPU.

[@JoaoCarlos-bp8ip]

@@leonvanzyl maybe the lack of nvidia gpu? mine is an AMD, because here I didn´t get it to answer so far.

[@JoaoCarlos-bp8ip]

I´ve found the issue. It was the way I wrote the ollama server address, it worked 127.0.0.1:11434/ this way instead of localhost

[@SyntharaPrime]

It is amazing. Great effort and very usefull. Thanks alot

[@youriwatson]

Thanks for the video Leon! When making chatbots for clients, do you encounter any cybersecurity concerns? for example, i could imagine clients worrying about using a third-party script on their website, because it could be vulnerable to XSS attacks, user privacy issues etc. Would like to hear your (or someone else's thoughts on this)

[@leonvanzyl]

Good question. I've had to create some custom integrations for clients where all sorts of security and Auth systems are in place. I feel that Flowise provides enough flexibility in the platform, especially when using their APIs and the new SDKs, to build whatever you want within the requirements from the client.

[@msclt1982]

Thank you so much Loe, You videos are crisp and clear. I have created a chatbot, as shown in the video, using LLaMA and configured it on my VM. The chatbot has memory and is responding to questions as expected. However, the problem I am facing is that the bot takes a lot of time to load and respond to each query. I have already built a chatbot with OpenAI, and while it was a different approach, I didn’t encounter this kind of slowness. Anybody faced this issue? Do you have any solution for this?

[@leonvanzyl]

Your VM might need some more horsepower

[@msclt1982]

@@leonvanzyl wow I didnt expect a quick response from you. Thats great. :). Anyway iam using an Azure VM, and already checked this approach, and the bot functionality improved after increase the vm cpu and ram. Thank you so much.

[@jonathanmisaelmusser8090]

Great video! But I have a problem, the chatbot doesn't always respond, or I see the response when I close and reopen the window, what could be happening?

[@pategoubusinesscenter1307]

thank you for this video, please hox can i add it to a website ?

[@leonvanzyl]

I have a video on my channel that shows how to add Flowise chatbots to websites. Obviously you won't be able to use a local instance of Flowise for that, so you could either self host Flowise or use their cloud service. Instead of using Ollama, use the Groq Chat node instead.

[@ageell2004]

thanks for the hard work. however how do we able to "export" the chat bot into a single site instead have to start with the chat bubble?

[@КравчукІгор-т2э]

Thanks for the content!

[@Karacho-u7v]

Somethings missing from the process "AppData\Roaming pm ode_modules\flowise\dist\commands\start.js: Cannot find module 'langchainhub'", how to get? Pip doenst work on windows.

[@sridhartn83]

Subscribed and Liked. :-) This is really helpful and simple, something I really wanted to learn and set up. thanks a lot, I have set up the chatbot for my pdf repository, but is there a way to pull just the chatbot something like ollama openwebui chat, what I mean is- once we set it up in flowise, as I see I open locathost on port 3000 it takes me to the flowise page and then I click on the chatbot , the whole chat flow is shown there then I have to click chat icon to initiate a chat , is there a way to setup an url:port I can put on browser and it lands me directly on the web UI showing me this flowise chatbot that I configured.

[@sridhartn83]

NVM I got the url , in the settings option in chatbot also other APIs, great thanks again. Really appreciate you for making this video.

[@karthikeyakuncham6929]

Hey nice video leon....but in my Rag chatbot I am using a tool agent and I need to pass the context which I have embedded in vector db in chat prompt template so that the agent can analyse whether the question is related to the docs provided or context provided ....how can I pass it?....i just want my chatbot not to give answers for unrelated questions. can you please help me with this

[@PIOT23]

Brilliant video as always 👏🏼

[@researchandbuild1751]

I installed flowise using their own documentation and during the NPM install it has a TON of warnings about unsupported packages lol!

[@Col-pd2zd]

Is it possible to use LLaVA with ollama to have the client upload a photo and the model can view it? I know it does'nt have the option, like how openAI has.. but maybe you know a work around?

[@tommoves9935]

Thanks for the video and setup. With OpenAI, I get it to run great. However... when I try to use the Conversational Retrieval QA Chain or Conversation Chain node in combination with ChatOllama, I get a "fetch failed" error (whatever model I use). When I run Ollama (+ model) with a regular LLM Chain node, it functions as well - but obviously no RAG here. Too bad... Has anyone any idea?!? I would appreciate that very much!!!

[@jawwadhussain8457]

you made my life complete :P with love

[@leonvanzyl]

Hehe, glad I could help 😀

[@APOLOVAILS]

looks really nice but coulndnt instal flowise, i get " Cannot find module 'colorspace" error when try to start npx. hope i will find some help on the web. thanks anyway !

[@showcaseshot]

when i type something in the chat box i get the error Fetch Failed. Do you know how to fix this issue? I searched the web got a suggestion to reinstall ollama but still the issue persists.

[@TabindaQudrat]

Getting same issue. do you find anyway to resolve it?

[@soundlab4831]

Hi, in Base URL enter 127.0.0.1:11434

[@stanTrX]

Thanks, flowaise looks good but is it free?

[@leonvanzyl]

Yes, it's open source and free to use

[@warrior-dl9vn]

I get fetch error for any prompts in the chat? what could be the issue?

[@damirrekic4271]

I had this issue too, but after upgrading the node to version 20.18 , it works properly

[@univer6979]

@@damirrekic4271 you are right

Awesome! How can we publish this on our website? Hosting it?

[@leonvanzyl]

Hey there! I have plenty of videos on my channel related to deploying Flowise and embedding it into websites. Ollama is meant for running the models locally, but in the cloud you could use Groq to run open source models for free. Again, I have videos on using Groq as well 😊

[@ShaunyTravels.]

Knew if there is one place to come to get help its here !! Leon need help connecting a multimodal model from ollama in flowise for text and images do you have a video about that I looked but can't find one 😢

[@leonvanzyl]

Hehe, my PC won't be able to run a model with image capabilities 😊. Will see what I can do once the larger Llama 3.2 models become available

[@jawwadhussain8457]

i have a issue with MSSQL Server i have trying to get data from another mssql server machine kindly make a small video of this

[@RedCloudServices]

How can you setup a vision model to read from a pdf.

[@kbtvn]

OpenWebUI or Koteamon can do the similar thing?

[@leonvanzyl]

I doubt you can build multi-agent apps, or agentic rag solutions, with those platforms 😊 . HEHE, seriously though - Flowise allows for some way more advanced AI solutions than those interfaces.

[@PrensCin]

how can we ollama 11b model move avilable models?

[@Aibots777]

Do i have to add doc to loader? Why can’t i point to folder to load ? Cuz i got 16 tb file that will not fit

[@leonvanzyl]

If I'm not mistaken you can use the folder or file loader for that. That is a massive file though, so you might want to clean up that data first, to text only, before uploading it.

[@tecnopadre]

Well, for files with 1 KB it's awesome. When you load files with let's say 20 MB, the embedding doesn't work. Maybe it's the Ollama computer. Trying to find out.

[@leonvanzyl]

Interested. I'll actually follow up on this. I've personally loaded some massive docs IRL and haven't experienced issues. Did you remember to split that doc?

[@tecnopadre]

@@leonvanzyl Yes I did. I alwasy do with REcursive Character Splitter (Showing 1-50 of 124 chunks 76.210 characters). And I had to use For the Upsert, Voayage embeddings and Faiss. I cuoldn't make it with Ollama Embedding because my PC CPU and memory couldn't handle it. Although it has a NPU. This is something Ollama has to work on. How to run Ollama with NPU instead of CPU and Memory.

[@alexandruscinteie7449]

@@leonvanzyl I have the same issue as the original commenter. I have uploaded a 600 something kb pdf and it can't return any answers. I followed your instructions to the letter. The only difference is that i have used the smaller llama3.2:1B model. All i get is "Hmm... I'm Not sure" answer from the bot. Any advice would be greatly appreciated! My intention is to create a document chat bot for documents to summarize and extract data from larger files.

[@karthikeyanvelayutham2817]

Hi Leon, I was try your way to create AI chatbot, the only problem I'm not able to load a document bro, I used win machine. is there a way that you can help here?

[@leonvanzyl]

I also use a Windows machine. What issues are you experiencing? You should check out my Document Store video if you're new to using Document Stores 🤞

[@donatocontreras7073]

How can I use more than two document stores at the same time? If I can only use one document store card (vector)?

[@leonvanzyl]

Absolutely! I would use a tool agent and attach two retriever tools.

[@donatocontreras7073]

@@leonvanzyl please! Could you show in a video how to do it?

[@mishaelo689]

Is it possible to host ollama on coolify?

[@DarkLineSnes]

why do we need this vector thing? i didnt understand what's it for

[@leonvanzyl]

The vector store is where the knowledge base is stored.

[@longsyee]

Just wondering when will the next update for llama vision that can read images 😆

[@leonvanzyl]

I'm keeping an eye out 👀. Look forward to putting it to the test.

[@AlphaMoury]

Is it possible to load multiple models in parallel using Ollama and Flowise?

[@leonvanzyl]

Sounds interesting. What would be the use-case?

[@santhoshkrishnan6269]

getting fetch failed error. please help

[@TabindaQudrat]

Did you resolve the error as i am also getting the same error.

[@soundlab4831]

Hi, in Base URL enter 127.0.0.1:11434

[@colabwork1910]

I don't when I get the registration link from Flowise. It almost 15 days now since I requested for account registration. Any suggestion, please?

[@leonvanzyl]

You can avoid the waiting list if you use my link apparently.

[@univer6979]

How can I set it up on a workstation which doesn't have internet? How to download the models for offline installation?

[@leonvanzyl]

You'll need an internet connection to download the models from Ollama. There's probably a way to drop a model into the Ollama folder, but I haven't tried that yet

[@univer6979]

@@leonvanzyl I copied the files to the offline pc and restarted, it worked. Thank you. One more question - is there a way to install flowise offline? File copy isn't working in case of flowise

[@BirdManPhil]

its not free. you get a 14 day trial then you pay $35/mo for 10k predictions/mo to use your agent. am i wrong or does it clearly state that in the PRICING page

[@leonvanzyl]

Nope, it's free and open source. Follow the instructions in the Flowise section of the video to set it up on your machine. What you're referring to is their managed cloud service which is definitely not needed.

[@Opeyemi.sanusi]

I can't create a document store. It says created but it's not created

[@leonvanzyl]

Haven't had that issue myself. Maybe someone in the comments knows. Try to update FW perhaps?

[@Opeyemi.sanusi]

@@leonvanzyl other people on Github had the same issue but I was able to fix it I switched to Node.js version 18.20.4 and Instead of installing it in my Documents or Downloads folder, I created a new folder directly in my root directory at ~/User/AI/, which helped avoid permission issues. After navigating to the folder in terminal, I ran npm install -g flowise and that did the trick!

[@shayanumar6824]

Can we integrate these chat bots in our own websites?

[@leonvanzyl]

Absolutely! Check out the web embed video on my channel

[@shayanumar6824]

​@@leonvanzylone more question it is taking longer time to generate the response is it dependent on my PC.

[@shayanumar6824]

@@leonvanzyl I have 8gb ram but still the response is taking a lot of time. is 8 gb sufficient or not.

[@RameshBaburbabu]

Awesome, and thanks for uploading. I would like to know how we can implement voice input and receive voice output from my internal RAG system.

[@pategoubusinesscenter1307]

interessing me also

[@leonvanzyl]

I don't think this is possible within Flowise itself. Would be awesome thought.

[@muraliytm3316]

Hello sir I got the following error while loading the database Oh snap! The following error occured when loading this page. Status: 500 Error: documentStoreServices.insertIntoVectorStore - Error: documentStoreServices._insertIntoVectorStoreWorkerThread - Error: Could not import faiss-node. Please install faiss-node as a dependency with, e.g. npm install -S faiss-node. Error: Could not locate the bindings file.

[@DanielLaLiberte]

I got a similar error: Error: documentStoreServices.insertIntoVectorStore - Error: documentStoreServices._insertIntoVectorStoreWorkerThread - Error: ENOENT: no such file or directory, mkdir 'C:\Users\danie\sentience\"C:\Users\danie\OneDrive\Desktop\faiss"' It looks like the path is appended to the current folder where I built Flowise, in the sentience folder. Yes, I just replaced the path, including the double quotes, with 'faiss' and it worked!

[@muraliytm3316]

@@DanielLaLiberte Yeah you can fix it by installing flowise in docker. Later if you use ollama in workflow, change baseurl, host.docker.internal:11434

[@ashypeshy]

I am on the waiting list to get acces.

[@leonvanzyl]

If you use my affiliate link you will skip the waiting list

[@meuchat_oficial]

And about to calling a function?

[@leonvanzyl]

Yes, these models support function calling 💪

[@inout3394]

Epic thx

[@leonvanzyl]

You're welcome 🤗

[@大支爺]

sponsored AD

[@leonvanzyl]

Nope. I did not get paid by Flowise, Meta or Ollama for this video. It wasn't created with their knowledge or permission either. I'm always blown away by comments like this. These videos are a lot of effort to create. Dude, I would happily accept a sponsorship deal if it means funding the channel, allowing me to create more videos. So even if the videos WERE sponsored, decide whether the content is for you and move on.

[@pongtomar]

Hi Leon , Thank you for sharing this clip. It is very useful for me. Now I was encountering the issues after upsert at Flowise even I was using Ubuntu 24.04 and having /home/myname/vector path" 0Error: documentStoreServices.insertIntoVectorStore - Error: documentStoreServices._insertIntoVectorStoreWorkerThread - Error: Request to Ollama server failed: 404 Not Found". Please give me some advice. Banatus

[@kuochingliew]

i stuck at Faiss ..but got this error tus: 500 Error: documentStoreServices.insertIntoVectorStore - Error: documentStoreServices._insertIntoVectorStoreWorkerThread - Error: Request to Ollama server failed: 404 Not Found

[@leonvanzyl]

Did you set up Ollama? That message is saying that Ollama is not running.

[@TheCcamera]

@@leonvanzyl having the same issue. (posted above in the comments). Ollama is up to date and running fine otherwise. I also successfully upserted vectors to faiss with Ollama embeddings with small documents but if you using larger documents it fails constantly. As workaround is it possible to use different embeddings with Ollama models?

[@surbhosale]

getting the same error even with a small document (I am using faiss with ollama embeddings)