This video took some time to create and I hope you guys enjoy it! This is useful for those of you who want to make money with these solutions, or for those who simply want to enhance their own businesses with AI Assistants. Feel free to share your implementations in the comments and I will check them out!
I could watch your tutorial videos all day. I think the best I have run into. I can't believe you compressed that to under 30 minutes yet made it very easy to understand. Your tutorials are expanding my mind of what is possible today and easy it can be. The traditional way of getting this experience prior to GPT would have been hours upon hours of professional services from developers and designers
Thank you for creating this tutorial. I appreciate how much time and research has gone into growing the knowledge that you have shared. Your delivery is ALL business and NO fluff - and it is brilliant. Thank you again.
There are a lot of self claimed AI Experts out there, here is a guy who do not claim anything, and yet output one of the most structured video on the subject I've seen, Sir thank you !
Thank you, Leon! As always, your content is incredibly helpful, and I really appreciate the great videos. I have a couple of questions. We have several different setups, for example, one for flight and hotel booking, another for doctor appointments, and another for searching nearby bars. These tools could functions or classes which perform well on their own. If we want to connect these independent tools via FlowiseAI, how could we go about that? For instance, if a user asks about nearby bars, the chatbot should forward the request to the relevant tool and return the results. Is it possible to connect FlowiseAI with our custom tools in this way? I'd appreciate any suggestions you might have.
Can demonstrate how to use Flowise AI to connect to the internal PostgreSQL database and reference OpenAI's assistant API to analyze the company's internal data? Thank you!
Another excellent video and perfect timing. I'm using n8n instead of Make, and was wanting to get Flowise up and running this weekend to do something similar. Your video will help a lot.
Thank you so much for sharing your knowledge! Is it true that the chatbot code your Flowise chatbots use is for development only? Or can it be used for production after error handling is added? Also, might you do a video on production code error handling for Flowise chatbots?
Very good. You were able to send the data via webhook. But let's imagine the following scenario. You send an order code, the system searches the ecommerce database and returns the order data. How would you return the result to the bot?
Hi Lyon, I create the agent exactly like your example, but when I test it, it returns invalid input syntax for type uuid: "thread_YNsS6ZMR0iX3W3Suu83YLhCt" How do I fix this? In the openAi playground it works very well, but in flowise it doesn't.
Amazing Video Leon. Super illustrative. I have a question regarding the Assistants responses passed in this case to Flowise. Do those allow Dalle-3 images same as GPTs do? Many thanks in advance.
Thank you for the compliment. Assistant API solutions are seperate to GPTs. I think certain UA-camrs are confusing the market by calling Assistant API solutions custom GPTs.
Hi Leon. I am following your series. Awesome. What makes this way (integrating OpenAI Assistants) different to the methods you showed earlier in this "UA-cam Flowise AI"-Series? Why should I use this method and not the one you tought earlier? Is it only that I don't need the other services like Pinecone? So I am a bit confused where and how to start. Cheers, Karsten
You've hit the nail on the head. With the Assistant API you don't have to pay for seperate services (vector database and memory). There are disadvantages as well. Since all the services are with OpenAI, you are locked into their ecosystem.
@leonvanzyl Was it necessary to use Make in order to store the lead into Zoho? Couldn't you have sent the information directly to Zoho? Excellent illustration by the. way.
Leon, excellent video. Exactly what I was looking for. This is actually useful in real-world scenarios. Thanks. Use Case: I had a specific use case: Instead of Zoho, I want to connect to my custom Field Service business Application and say create tickets, check work order statutes etc. I have total access to this application [APIs and DB] and there are about 100 operations I want to perform on my application using LLM capabilities. Had a few questions. 1. What would be the best way to achieve the above integration. I guess we would need to use OpenA's function calling using Assistants. Will I need to add code/customize for each API? 2. I have heard about NL2SQL and Lanchain's SQL Chains. How do these all play together?
Thank you for the feedback, glad you found this helpful 😊. Your use case is definitely achievable by adding a few more custom tools. It's a bit too much for a comment, so please reach out if you need assistance. My agency's details are in the description.
Hello and nice video as always. A question comes to my mind about the prompts: In general, what is the practical purpose of telling the AI what it is (i.e. "You are a gym coach ....")? Does this mean that the AI has some algorithms and will somehow focus only on that part of its "brain" (data/knowledge) that is related to the defined role?
Defining the role changes the persona of the bot slightly. You don't have to specify it, but you might get some neutral and boring responses from the bot 😁
Amazing video Leon. You're a legend. Great explanation on all your videos. May I ask you something? I have tried to set up the assistant with Claude Haiku, but when it comes to send the lead, it does not do it (I am using Open AI Tool Agent). It answers everything and does everything needed. But when it has to send the information to the lead, it doesn't. Is it just ChatGPT that does it? Thanks!
Thank you! Only OpenAI and Mistral supports function calling. However, Claude also supports it now but I think the Flowise team is busy implementing that functionality.
Great video, thank you! I followed the steps and have the widget on my WP site, but every message says "failed to fetch". I have it hosted on a digital ocean droplet, and it does work on the cloud instance of flowise, but the embed isn't working. I didn't incorporate the lead gen, just the OpenAI assistant with a file retrieval. Any quick thoughts/fixes maybe?
Hi Leon! The content you provide is extremely helpful! I really appreciate it. I would like to ask you something, if you can help me (I'm a little confused with all the features). I'm building a Conversational Retrieval QA Chain. Inside the node -Additional Parameters- I defined a prompt of the task I ask the LLM to perform. I'm not focusing in allowing users to chat with the pdf; instead I want the user to provide a pdf, and based on that pdf the LLM generates some key definitions. I want to get these "key definitions" and then show then in my web app (using a different UI, not a chat). How can I "capture" the outputs that the node generates; and the post that outputs directly in my web app. So this is the flow in my web app: User uploads PDF > The app autogenerates some key definitions > The key definitions are showed in the web app (with a special UI style -like cards- and also interactive. If the user clicks in the "card" a larger explanation is showed I hope I've been clear in my doubt. Thanks for all you are doing!
Hi Franco, Excellent question. Unfortunately Flowise Embed (the chat bubble) won't be able to hand over information to your website. It's basically a seperate app that is simply layered on top of your web app, with no real way of interacting with the UI. I would solve your requirement in a different way. Perhaps your website could call the Flowise API instead, and then populate the form based on the API response.
@@leonvanzyl Thanks for your answer Leon. Yes, I thought I wouldn't be able use the embedding. However while interacting with the api, I'm quite confused in how to "capture" the responses in a structured way. I mean, how to control the responses from the LLM in a standard format; and the "paste" the responses to another service
Thank you for sharing this Leon! I replicated it as the video, but using n8n intead of make. I also tested a get_lead function data info from the data base and it works pefect. Thank you so much!
I think there has been an update in the platform that leaves a big question about the process. now you can not upload the document / file, for file search it asks to add vector store, how do you do that? or should you upload the document in cod interpreter files?
Thanks Leon, your video explanation was excellent. Could you consider creating a video on how to develop multiple digital assistants? Alternatively, a tutorial on integrating an assistant with WhatsApp or other social media platforms would be greatly appreciated. Your insights and guidance would be incredibly valuable in these areas.
Hi Leon, I've followed your instructions for the OpenAI Assistant, but the models you mention are not available anymore so I've chosen 4o mini (tried others too). When I say "hi" in the chatbot then I have a weird message: "Error: 404 No thread found with id 'defab' [+ series of numbers]." What should I do to fix this? Thanks!
Maybe it's time for me to create an update to this video. The Assistance API had changed a lot since then. Try using GPT 4o instead. Mini might not be intelligent enough.
Hi Leon, Ihope you are doing great! This tutorial is the closest to a masterpiece so far! Thank you!! If I wanted to move the file loading out from openai, and doing it in flowise to make it dynamic via an API call, will it work if I create a separate RAG flow and then add it as a second tool in this flow? Thank you!!
Hey Ricardo! I don't think there is a way to upload files to Assistant API via Flowise. I might be wrong though. You might be better off just creating your own tool agent. ua-cam.com/video/Dinzg_dGfqU/v-deo.html You can use the Flowise API to update the vector store. As another technical alternative, you can call the Assistant API directly to upload / manage files. You'll need to reference the OpenAI docs for this one.
I understand. Better not to mix things. If I want to manage from Flowise I should better go Flowise all the way and create my own agent with similar functionallity. Thank you!!
Thanks for the wonderful tutorial. Can we use this tool to also capture documents and pass them to the OpenAI assistant? We are looking to analyze files that a user would submit.
After uploading flowise to render do we have to open flowise locally to access the cloud version because mine url of render is not working without opening flowise locally
Hi there. OpenAI Assistant provides RAG solutions out of the box. You can upload files to it's knowledge base, or provide tools (like API integration). The best way to use these would depend on your application. If you simply need a customer chatbot that is trained on your business information, then you can just upload files to the knowledge base using the OpenAI dashboard.
when i try the webhook part testing flowise i get a reply from my agent saying Error: Error executing tool. Tool: add_trial. Thread ID: thread_k3NsSqn9FNpLM2xPvjmRDFcl. Run ID: run_LfC7g8p2lIy3l76NIUAPmIh8 what does this mean? it isnt linking to the webhook so i am stuck currently. any help please?
Very helpful video Leo. Amazing tool. I have two questions. I build on react, is there a good case, where it is ok to embed in client directly, I see that the flowise playground link is exposed? This leads to my second question, when I host the flowise online, should I not protect the playground from access, or anyone can open the deployed link and make changes?
Nice work, video helped me a lot compared to other videos I have seen! Is it possible to change the opening sentence of the chat bot? It's too much generic the sentence - Hi there! How can I help?
Thanks Leon - amazing - for some reason im getting a flowise error when I test the assistant. It works ok on the open ai side but has an error of chat flow not found when testing on flowise ? Any ideas ?
@@leonvanzyl thanks for the lightening fast response the error related to a chat flow and upon checking there was no chat flow only an assistant :-) Are you open for a call ?
Hey Leo, I love your videos. Is it possible to use AzureGPT as OpenAI Chat Model instat of ChatOpenAi? And how many Assitants can I add maximum to the Flow?
Thanks for the feedback 🙏. This is a service from OpenAI, so unfortunately you can only use their models. I think you can only have one agent per Chatflow, but you can get Chatflows to call other Chatflows using custom tools and Flowise APIs.
That's always a good question 😁. My advice is to not underestimate the amount of effort needed for things outside of the chatbot itself. I have just finished work for a client on implementing exactly this type of assistant on their website, and had to spend a lot of time setting things up on THEIR hosting servers and website.. which took an unforseen amount of time. I would not think it's unreasonable to ask anything between $500 - $1000.
Thank you! Assistants API is a service offered by OpenAI and comes with a built-in memory solution using Threads, so unfortunately you cannot swap out the memory and vector solutions. In that instance you are better off just building your own agents - I have a series on Flowise that shows you how to do this.
hello again. I love this even more in testing! Running into a slight challenge when calling the FW API inside another bot builder - ive set the api call to raw to capture last user question but as soon as the FW api responds it stops the chatflow - any thoughts on how to keep the FW chatflow "alive" so I can keep just using the FW chatwindow...thanks!!
Glad you like it! You'll have to get creative with the sessionId when using the API endpoint. If you don't pass a sessionId, then you're effectively creating a new thread / conversation with each API call.
Thanks for the video. This is really helpful. A couple of questions - 1. Is there a way to customize the generic greeting message that the bot presents? 2. Is there a way to get user feedback on whether they liked the bot answer? Thumbs up/ Thumbs down. I could possibly have asked these questions on Discord but the link is not working for me.
Thank you for the quick response@@leonvanzyl. Will look up the other video. In the absence of user signals - Is there any other way to figure out how Assistant is doing? How do we know where and when to finetune?
Many thanks for you job! Helps to understand all non code tool. It’s any chance to show how can be integrated gpts with ms teams as help channel in company?
o meu sempre retorna esse erro Error: Error executing tool. Tool: add_lead. Thread ID: thread_9VvZoXrggIAa31jsCgqcByjw. Run ID: run_5VI61i5hsaPRZ2fGaFv3H6mY
Hello @leonvanzyl, How would I save the chat history in flowise? I mean after deploying the model on a website (as shown in your previous videos), I am interested in keeping a record of all the conversations a client has made with the bot and emailed to me along with the lead notification. Is that possible ?
Great tutorial Leon, just discovered these this weekend. A quick question, I installed flowise on my MacBook and it works great. When I embed it onto a web page, also on my localhost, it works perfectly but there are no avatars / icons. They are there inside flowise but not on the webpage after embedding it. Any suggestions?
Hey, you're welcome 🤗. Are you referring to the Assistant Avatar? I also noticed that it's not actually used anywhere. I'll ask the Flowise team about this.
Thanks for the video! I have a question. May I know the difference between extracting information from PineCone database using index and extracting information from the knowledge files in the OpenAI assistant? I think previously without OpenAI assistant I can also extract data and add custom options to the chatbot?
You're welcome. The difference is that with OpenAI Assistants, you do not need to use 3rd party services, like Pinecone, to store your knowledge base. It's built into the Assistant API platform.
@@leonvanzyl I tried this with the open ai assistant. If multiple files are loaded, I find that it will only focus on one document and ignore the remaining documents. In other words, it won’t search through all uploaded files in the knowledge files. But using pinecone, all files contents are indexed for search.Do you encounter this problem?
@leonvanzyl I'd really like to learn how to add Speech to text (for user input) and Text to speech (to talk back to the user) to make interaction with the tool way easier. I guess this is custom tooling, do you happen to have a tutorial on this? With openai's whisper and openai's audio speech endpoints this should be possible, however I'm to much of a "no-code" guy to understand how to do it without instructions.
I'm using the endpoint provided by Flowise to integrate the workflow with my own chatbot. Is it possible to pass a parameter in the body that contains a URL for a doc to upload and to be able to ask it questions?
Same thing. A Conversational Retrieval Agent is also known as an AI assistant. The Assistant in this video is referring to the Assistant API provided by OpenAI. It's an Agent that is fully hosted by OpenAI. It's still nothing more than a Conversational Agent in the backend.
Use Odoo for my B2B ERP, where the shop is hidden unless logged in. Want to create a Conversational QnA but define a rule to only access that customers data. Where/how would you recommend the rules to be implemented for the postgresql request? Haven’t found any documentation regarding and hopped maybe you had a source document you could recommend, or a video I missed.
Awesome tutorial. Just wanted to ask since its uses Assistants API which is paid, what if the user chats about other stuff like hey buy my rayban glasses etc. How do we keep it limited so that it doesnt waste credits?
Thank you! I probably wouldn't use Flowise for those integrations at the moment. These apps have incredibly tight timeouts unfortunately, and the assistants API is quite slow on the responses at the moment. Once this improves, I'll definitely create a tutorial on it.
@@leonvanzyl thank for the reply. What cost effective options do we have to use these assistants on whatsapp, insta and facebook? Voiceflow will be expensive I think
Thank you for sharing. I encountered an issue during testing where the add_lead function doesn't record Chinese characters when the user's dialog text is in Chinese. Is there a solution to this?
would be nice if flowise would expose endpoints for each created agent so we could use them otherwise. just came around this nice tool. would make agent creation for external use much easier.
Great video, I was trying to run it lately but I noticed the Assistants API's latency is pretty bad - sometimes it takes up to a minute for a response to get back. Do you know if there's a way to improve it or is it going to scale up soon? Thanks
Hi Leon - ive have the chatflow disappear now several times after updating in make. Are there any known issues with flowise ? The assistant still exists in open ai
Do you have a complete paid course where you can learn all the tools that flowise has because they talk about very few and I know that there are many more that are not found on UA-cam
Hi Leon, thx for this cool Video. I really like the way you describe the entire process making this chatbot with Make and Zoho CRM. But I couldn't make Zoho CRM working, because I do not get a list of Values (connection is valid and verified!) to map the values from the Webhook. If I try with Hubspot, it works. Any Idea, why I do not get the list of possible values for the mapping in the Zoho CRM Module?
You're welcome 🤗. It's a bit hard to say given the informations, but the ZOHO CRM module should automatically provide all available fields. Perhaps your ZOHO connection isn't set up correctly.
@@leonvanzyl I did setup a second Account on Zoho CRM and also I used a Windows 11 Environment and Edge instead of MacOS Monterey and Safari to test it. I got the same bad result - can't se any fields like Company, Last Name, etc. Not before nor after I established the connection to Zoho CRM as shown in your Video. Are there any API values that one has to setup after creating a Zoho CRM account?
I have the same problem. It only shows me the basic fields, and is missing many of the other ones. The one I need is "Description". If you find a solution or have any alternatives, let me know :D @@MisterOcean5
I'm trying to connect Hubspot with a "conversational REtrieval QA Chain" but I can't figured out. I'm seeing the only way is with "OpenAI Assistant"? Thank you
What's the error / issue that you're experiencing? I wouldn't think it would make a difference, since the agent and OpenAI Assistant make use of the same tools and document loaders.
Perfect!!! Just only an input. When I create the under make the zoho crm It is created correctly but the windows where you mach the fields doesn´t appear, so strange.
It's the EXACT same process described in this video. Instead of ZOHO CRM, connect Calendly instead. Adjust the custom tool and prompt accordingly. In fact, you could add a second custom tools for getting available dates / times as well.
@@leonvanzyl aha yes makes sense thanks! Will see how calendly looks inside this…Yes I want 2 custom tools…one to check times then send to CRM…this totally rocks…you’ve saved me a ton of time as was on my list for the week!! 👏👏👏👍👍👍
You can use Open source models to achieve some of this functionality, but OpenAI is the leader when it comes to function calling at this stage. Check my videos on calling Open Source models 😁
Doesn't exist the Actions functionality inside OAI Assistants that could replace the Custom Tool used on Flowise Node? Probably the code will not be same JS based as the example too.
Hi Leon. You are a legend. I have learnt a lot with you! I am loyal subscriber!! Could I ask you someting? I followed your tutorial and it worked with your data. But when I try to write down mine, it gives me this error. Error: Error processing thread: requires_action, Thread ID: thread_T3K40MVIeeAzu0ET7OUAVPKp, Run ID: run_Lzs714A7v078l2pzc5FmiyKB, Tool: add_lead It is exactly the same as you did, I just added two more functions besides name, phone and some others. Do you know why this is? I have been trying the whole day to fix it and it is really driving me crazy :(
@@leonvanzyl I could fix it. It's taken me more than 6 hours to figure it out, but it's alright now. Thanks for the response anyway! Looking forward to your next videos!!
The template works wonderfully. However, my bot disappears in the chat flow of Flowise after about 2 hours and is not permanently saved. What could be the reason for this?
The free account does not have permanent storage. Check out my deployment video to see how to set up a paid account with storage. If you're using this for business or clients, the $7 p/m is well worth it.
Custom GPTs are accessed by the ChatGPT website and are basically custom versions of chatGPT. The Assistants API is not related to Custom GPTs and does not have the same behaviour or functionality. Instead, it a framework that developers can use to build advanced agents / chatbots. As for WhatsApp/ Intragram integration, it's way too complex to answer in a comment. In a nutshell, you need to familiarise yourself with each of those platforms' developer tools and how their webhooks work.
Is it possible to use an Agent that can call tools just like in this example but also perform lookups on a vector database? Kinda like a mix of Conversational Retrieval QA Chain and OpenAI Assistant.
Yes, that is exactly what the assistant does. Simply upload document to its knowledge base and the agent will automatically figure out when to use the knowledge base (a vector store) or call tools.
@@leonvanzyl But that's quite limiting compared to a classic rag with a vector database where I can store vector representations of any kind (and amount) of data, be it code or text split in the way that I'd consider best suitable for it (chunk sizes and so on).
@@RobertoFabrizi yeah, that's one of the disadvantages of OpenAI Assistants. Building your own solution gives you way more control and you are not locked into the vendor. Trade-offs 😁.
@@leonvanzyl If I may ask one last question on the matter (thank you so much for your time!!) how would you suggest that such use case be implemented? Leveraging the ReAct agent to invoke che custom tool?
@@leonvanzyl No, I mean the Pop-up chatbot on my wordpress website. The problem is that in the mobile version I can scroll through the messages with my finger but in the desktop version I have no way to scroll up or down in the chatbot. Even if i hover over the chatbot with the mouse pointer, the entire website scrolls but not the chatbot
Hi leon! thanks for sharing such an amazing content with us... I was trying to integrate a flowise powered chatbot on my worpress website and followed all the steps mentioned in the video but getting error... the error is about rule 941110 - reason XSS Filter... can you please help in resolving it, I ll really appreciate it...
Don't you just hate it? Following tutorial step by step and when it gets to testing it's not working. Set up the assistant, uploaded the file with company info and when asking questions (which are included in the document) it tells me that it couldn't`t find the answer in the document provided. Checking with Open AI looks like the document is uploaded and tokens getting added like it is working but I get only negative answers.
@@leonvanzyl In Open Ai/Assistant at the bottom of the page is a field called API Version with an option to switch to v1. If we switch to v1 will it have an effect?
@@leonvanzyl This is useless man. I deleted everything and started from scratch 7 times. Watched your video 20 times just in case I missed something. It is not working. In Playground I get the right answers but in Flowise tells me it can not find the specific information in the document provided and I asked the same question. Is there any other way to do this?
Noticed that too sometimes. I've reached out to Flowise so they can hopefully address it. The Assistants API natively returns the citations on the responses, and it's up to the consumer (Flowise in this instance) to remove the citations. Hopefully OpenAI will parameterize that functionality soon.
@leonvanzyl I've noticed that Flowise has some output parsers, but none of them allow me to run custom JavaScript code to clear those entities out. If you receive feedback from them, please let me know. I appreciate it! ;)
just wonder, after crating the assistant on Flowise, can I add functions through the Openai website? or stll have to add through flowise? it would be absolutely great if we could do modification later directly on their openai page
Keep in mind that the Assistant is a RAG chatbot. I assume you're asking if you can add more than one agent to the same chatflows. I suppose it is possible, but it will involve calling different chatflows using they're APIs.
@@leonvanzyl i make a chat bot which speak on my own information (like in u RAG tutorial) And now i want to add booking function into this bot. And i have second question, how i can make the text bold in my bots response (like on u website)
@@leonvanzyl We are using 1.4.3 and do not see the Open AI Assistant under agents. Is there some trick to get this working? Thank you for your amazing videos. They have been so helpful.
thank you for your video, but I don't know why I make step by step like you do, but at 11:00 when I chat Hello, it show an errror Message like this: "Error: Error processing thread: failed, Thread ID: thread_mdI12dqWexCGUXv6d1ZXTHVB, Run ID: run_fmjL58EGTobM06y6IGbE0orB, Status: failed". Can you help me to resolve the problem. Thank you so much !!!
This video took some time to create and I hope you guys enjoy it!
This is useful for those of you who want to make money with these solutions, or for those who simply want to enhance their own businesses with AI Assistants.
Feel free to share your implementations in the comments and I will check them out!
So good! Knowledge leap - thanks Leon!
I could watch your tutorial videos all day. I think the best I have run into. I can't believe you compressed that to under 30 minutes yet made it very easy to understand. Your tutorials are expanding my mind of what is possible today and easy it can be. The traditional way of getting this experience prior to GPT would have been hours upon hours of professional services from developers and designers
Thank you for this amazing comment 🙏.
You're right, this was unthinkable a year ago.
Thank you for creating this tutorial. I appreciate how much time and research has gone into growing the knowledge that you have shared. Your delivery is ALL business and NO fluff - and it is brilliant. Thank you again.
Thank you! This means a lot.
There are a lot of self claimed AI Experts out there, here is a guy who do not claim anything, and yet output one of the most structured video on the subject I've seen, Sir thank you !
Thank you Anthony!
This means a lot. I'm over listening to posers myself.
Thank you, Leon! As always, your content is incredibly helpful, and I really appreciate the great videos. I have a couple of questions. We have several different setups, for example, one for flight and hotel booking, another for doctor appointments, and another for searching nearby bars. These tools could functions or classes which perform well on their own. If we want to connect these independent tools via FlowiseAI, how could we go about that? For instance, if a user asks about nearby bars, the chatbot should forward the request to the relevant tool and return the results. Is it possible to connect FlowiseAI with our custom tools in this way? I'd appreciate any suggestions you might have.
Nice work Leon. Thanks for taking the time to make this video!!
You're welcome 🤗
Can demonstrate how to use Flowise AI to connect to the internal PostgreSQL database and reference OpenAI's assistant API to analyze the company's internal data? Thank you!
@leonvanzyl THANK YOU SO MUCH. I watched 25 videos to find what I needed. Your video is incredibly helpful. Thank you for your time and effort.
Thank you very much for the feedback 🙏
@@leonvanzyl Any tips on how I might push an entire summary of a chat to a CRM?
Another excellent video and perfect timing. I'm using n8n instead of Make, and was wanting to get Flowise up and running this weekend to do something similar. Your video will help a lot.
Seems you and I are usually in sync 😁. Glad this could help in some way
Thank you so much for sharing your knowledge! Is it true that the chatbot code your Flowise chatbots use is for development only? Or can it be used for production after error handling is added? Also, might you do a video on production code error handling for Flowise chatbots?
You're welcome 🤗.
You can definitely use these solutions in production. You should always consider exception handling in your solutions.
Very good. You were able to send the data via webhook. But let's imagine the following scenario. You send an order code, the system searches the ecommerce database and returns the order data. How would you return the result to the bot?
Hi Lyon, I create the agent exactly like your example, but when I test it, it returns
invalid input syntax for type uuid: "thread_YNsS6ZMR0iX3W3Suu83YLhCt"
How do I fix this? In the openAi playground it works very well, but in flowise it doesn't.
This is brilliant, exactly what I was looking for, thank you so much!! More power to you bro 👍
Thank you 😊
This video is amazing. It's exactly what I wanted to achieve. Thank you very much Leon for the clear, step-by-step guide.
You're welcome 🤗
As usual, this is an excellent video with precisely the information I needed. Thank you, Wayne
Glad it was helpful!
As always, to the point and excellent example to get started!!! You and Flowise are simply great!! ❤
Thank you!
Amazing Video Leon. Super illustrative. I have a question regarding the Assistants responses passed in this case to Flowise. Do those allow Dalle-3 images same as GPTs do? Many thanks in advance.
Thank you for the compliment.
Assistant API solutions are seperate to GPTs. I think certain UA-camrs are confusing the market by calling Assistant API solutions custom GPTs.
Hi Leon. I am following your series. Awesome. What makes this way (integrating OpenAI Assistants) different to the methods you showed earlier in this "UA-cam Flowise AI"-Series? Why should I use this method and not the one you tought earlier? Is it only that I don't need the other services like Pinecone? So I am a bit confused where and how to start. Cheers, Karsten
You've hit the nail on the head. With the Assistant API you don't have to pay for seperate services (vector database and memory).
There are disadvantages as well. Since all the services are with OpenAI, you are locked into their ecosystem.
Amazingly pedagogic, just perfect!!!
I had to look up the word pedagogic!
Thank you 😊
@leonvanzyl Was it necessary to use Make in order to store the lead into Zoho? Couldn't you have sent the information directly to Zoho? Excellent illustration by the. way.
The issue is ZOHO OAuth authentication process. You wouldn't be able to manage ZOHO access tokens in Flowise.
@@leonvanzyl Thanks for the explanation. If you have time, will you do a video doing the same using Airtable?
Thanks! Very useful demo about integration FlowiseAI with OpenAI Assistants (custom knowledge base)! And great bonus about functions call
Glad you enjoyed it 😁👍
Leon, excellent video. Exactly what I was looking for. This is actually useful in real-world scenarios. Thanks.
Use Case: I had a specific use case: Instead of Zoho, I want to connect to my custom Field Service business Application and say create tickets, check work order statutes etc. I have total access to this application [APIs and DB] and there are about 100 operations I want to perform on my application using LLM capabilities.
Had a few questions.
1. What would be the best way to achieve the above integration. I guess we would need to use OpenA's function calling using Assistants. Will I need to add code/customize for each API?
2. I have heard about NL2SQL and Lanchain's SQL Chains. How do these all play together?
Thank you for the feedback, glad you found this helpful 😊.
Your use case is definitely achievable by adding a few more custom tools.
It's a bit too much for a comment, so please reach out if you need assistance. My agency's details are in the description.
Hello and nice video as always. A question comes to my mind about the prompts: In general, what is the practical purpose of telling the AI what it is (i.e. "You are a gym coach ....")? Does this mean that the AI has some algorithms and will somehow focus only on that part of its "brain" (data/knowledge) that is related to the defined role?
Defining the role changes the persona of the bot slightly. You don't have to specify it, but you might get some neutral and boring responses from the bot 😁
Amazing video Leon. You're a legend. Great explanation on all your videos. May I ask you something? I have tried to set up the assistant with Claude Haiku, but when it comes to send the lead, it does not do it (I am using Open AI Tool Agent). It answers everything and does everything needed. But when it has to send the information to the lead, it doesn't. Is it just ChatGPT that does it? Thanks!
Thank you!
Only OpenAI and Mistral supports function calling. However, Claude also supports it now but I think the Flowise team is busy implementing that functionality.
@@leonvanzyl Hopefully soon. Thanks!
Thank you Leon !! Do you know if its possible to embeed the chat to a Canva Website?...
Great video, thank you! I followed the steps and have the widget on my WP site, but every message says "failed to fetch". I have it hosted on a digital ocean droplet, and it does work on the cloud instance of flowise, but the embed isn't working. I didn't incorporate the lead gen, just the OpenAI assistant with a file retrieval. Any quick thoughts/fixes maybe?
Hi Leon, how do I get streaming responses using Flowise and OpenAI assistants, is that possible?
Hi Leon! The content you provide is extremely helpful! I really appreciate it.
I would like to ask you something, if you can help me (I'm a little confused with all the features). I'm building a Conversational Retrieval QA Chain. Inside the node -Additional Parameters- I defined a prompt of the task I ask the LLM to perform.
I'm not focusing in allowing users to chat with the pdf; instead I want the user to provide a pdf, and based on that pdf the LLM generates some key definitions. I want to get these "key definitions" and then show then in my web app (using a different UI, not a chat). How can I "capture" the outputs that the node generates; and the post that outputs directly in my web app.
So this is the flow in my web app:
User uploads PDF > The app autogenerates some key definitions > The key definitions are showed in the web app (with a special UI style -like cards- and also interactive. If the user clicks in the "card" a larger explanation is showed
I hope I've been clear in my doubt.
Thanks for all you are doing!
Hi Franco,
Excellent question. Unfortunately Flowise Embed (the chat bubble) won't be able to hand over information to your website. It's basically a seperate app that is simply layered on top of your web app, with no real way of interacting with the UI.
I would solve your requirement in a different way. Perhaps your website could call the Flowise API instead, and then populate the form based on the API response.
@@leonvanzyl Thanks for your answer Leon.
Yes, I thought I wouldn't be able use the embedding. However while interacting with the api, I'm quite confused in how to "capture" the responses in a structured way. I mean, how to control the responses from the LLM in a standard format; and the "paste" the responses to another service
Awesome! Thanks for this. Can ditch those over promoted chatbot designersat last yay!
Agreed
Thank you for sharing this Leon! I replicated it as the video, but using n8n intead of make. I also tested a get_lead function data info from the data base and it works pefect. Thank you so much!
love n8n, self hosted it??
That is awesome! Thank you for sharing.
yes, self hosted@@myandrobox3427
I think there has been an update in the platform that leaves a big question about the process. now you can not upload the document / file, for file search it asks to add vector store, how do you do that? or should you upload the document in cod interpreter files?
Thanks for bringing this to my attention.
I'll look into this and release an updated video if things have changed.
This space moves fast.
@@leonvanzyl Thanks Leon for the quick reply. I know, thinks change very quickly. Looking forward for you update.
Thanks Leon, your video explanation was excellent. Could you consider creating a video on how to develop multiple digital assistants? Alternatively, a tutorial on integrating an assistant with WhatsApp or other social media platforms would be greatly appreciated. Your insights and guidance would be incredibly valuable in these areas.
Thank you for the feedback 🙏
Totally agree, a tutorial on integrating an assistant with WhatsApp, Instagram or other social media platforms would be just great.
Fantastic! Any idea for spam protection or to avoid brute force attacks?
Yes, use the rate limiting settings as I've explained in the video 😊
Hi Leon, I've followed your instructions for the OpenAI Assistant, but the models you mention are not available anymore so I've chosen 4o mini (tried others too). When I say "hi" in the chatbot then I have a weird message: "Error: 404 No thread found with id 'defab' [+ series of numbers]." What should I do to fix this? Thanks!
Maybe it's time for me to create an update to this video. The Assistance API had changed a lot since then.
Try using GPT 4o instead. Mini might not be intelligent enough.
Hi Leon, Ihope you are doing great! This tutorial is the closest to a masterpiece so far! Thank you!! If I wanted to move the file loading out from openai, and doing it in flowise to make it dynamic via an API call, will it work if I create a separate RAG flow and then add it as a second tool in this flow? Thank you!!
Hey Ricardo!
I don't think there is a way to upload files to Assistant API via Flowise. I might be wrong though.
You might be better off just creating your own tool agent.
ua-cam.com/video/Dinzg_dGfqU/v-deo.html
You can use the Flowise API to update the vector store.
As another technical alternative, you can call the Assistant API directly to upload / manage files. You'll need to reference the OpenAI docs for this one.
I understand. Better not to mix things. If I want to manage from Flowise I should better go Flowise all the way and create my own agent with similar functionallity. Thank you!!
Is it able to send and receive images with the code interpreter?
I would say yes. Although not relevant to this video, it's definitely possible for code interpreter to load images and use Apis to generate images.
Thanks for the wonderful tutorial.
Can we use this tool to also capture documents and pass them to the OpenAI assistant? We are looking to analyze files that a user would submit.
You're welcome 🤗.
I don't believe it's currently possible in Flowise, but I'll definitely cover this scenario in a video once it is 👍
After uploading flowise to render do we have to open flowise locally to access the cloud version because mine url of render is not working without opening flowise locally
Whats the best way to connect this with Rag?
Hi there.
OpenAI Assistant provides RAG solutions out of the box. You can upload files to it's knowledge base, or provide tools (like API integration).
The best way to use these would depend on your application. If you simply need a customer chatbot that is trained on your business information, then you can just upload files to the knowledge base using the OpenAI dashboard.
Obrigado Leon! Acho que é isso que todos estavamos esperando.
You're welcome 🤗
I specified when assistant should use the function but it uses is it in wrong situations, e.g., when I ask for the name of the owner of the business.
when i try the webhook part testing flowise i get a reply from my agent saying
Error: Error executing tool. Tool: add_trial. Thread ID: thread_k3NsSqn9FNpLM2xPvjmRDFcl. Run ID: run_LfC7g8p2lIy3l76NIUAPmIh8
what does this mean? it isnt linking to the webhook so i am stuck currently. any help please?
Very helpful video Leo. Amazing tool. I have two questions. I build on react, is there a good case, where it is ok to embed in client directly, I see that the flowise playground link is exposed? This leads to my second question, when I host the flowise online, should I not protect the playground from access, or anyone can open the deployed link and make changes?
You're welcome 🤗.
Check out my video on deploying Flowise to Render. In that video I show you how to protect your Flowise instance with credentials.
@@leonvanzylthank you. Will do.
Nice work, video helped me a lot compared to other videos I have seen!
Is it possible to change the opening sentence of the chat bot?
It's too much generic the sentence - Hi there! How can I help?
Thank you! Glad I could help.
Check out my Web Embed video. You can change the welcome message using the configuration object.
@@leonvanzyl Web Embed video?
sand a url
Thanks Leon - amazing - for some reason im getting a flowise error when I test the assistant. It works ok on the open ai side but has an error of chat flow not found when testing on flowise ? Any ideas ?
What's the error?
@@leonvanzyl thanks for the lightening fast response the error related to a chat flow and upon checking there was no chat flow only an assistant :-) Are you open for a call ?
Hey Leo,
I love your videos. Is it possible to use AzureGPT as OpenAI Chat Model instat of ChatOpenAi?
And how many Assitants can I add maximum to the Flow?
Thanks for the feedback 🙏.
This is a service from OpenAI, so unfortunately you can only use their models.
I think you can only have one agent per Chatflow, but you can get Chatflows to call other Chatflows using custom tools and Flowise APIs.
Hi leon, very informative
question was that how much we should charge client for this same chatbot (excluding make/crm/open-api-key expenses) ??
That's always a good question 😁. My advice is to not underestimate the amount of effort needed for things outside of the chatbot itself.
I have just finished work for a client on implementing exactly this type of assistant on their website, and had to spend a lot of time setting things up on THEIR hosting servers and website.. which took an unforseen amount of time.
I would not think it's unreasonable to ask anything between $500 - $1000.
@@leonvanzyl thank u for being open 👍👍
Great video, is there any possibility of this storage being in Bubble?
Thank you!
Assistants API is a service offered by OpenAI and comes with a built-in memory solution using Threads, so unfortunately you cannot swap out the memory and vector solutions. In that instance you are better off just building your own agents - I have a series on Flowise that shows you how to do this.
hello again. I love this even more in testing! Running into a slight challenge when calling the FW API inside another bot builder - ive set the api call to raw to capture last user question but as soon as the FW api responds it stops the chatflow - any thoughts on how to keep the FW chatflow "alive" so I can keep just using the FW chatwindow...thanks!!
Glad you like it!
You'll have to get creative with the sessionId when using the API endpoint. If you don't pass a sessionId, then you're effectively creating a new thread / conversation with each API call.
Thanks for the video. This is really helpful. A couple of questions -
1. Is there a way to customize the generic greeting message that the bot presents?
2. Is there a way to get user feedback on whether they liked the bot answer? Thumbs up/ Thumbs down.
I could possibly have asked these questions on Discord but the link is not working for me.
You can customise the welcome message using the config object. Check out my embedding video 👍.
No.2 not possible at this stage. Would be cool.
Thank you for the quick response@@leonvanzyl. Will look up the other video.
In the absence of user signals - Is there any other way to figure out how Assistant is doing? How do we know where and when to finetune?
Many thanks for you job! Helps to understand all non code tool. It’s any chance to show how can be integrated gpts with ms teams as help channel in company?
You're welcome 🤗.
Will look into this.
@leonvanzyl also interested in how to enable this Flowise chatflow under MS Teams too.🎉
Ms teams would indeed be awesome. This man is the best in explaining flowise on the Internet. Probs man!
@@NS_Miata thank you!
o meu sempre retorna esse erro Error: Error executing tool. Tool: add_lead. Thread ID: thread_9VvZoXrggIAa31jsCgqcByjw. Run ID: run_5VI61i5hsaPRZ2fGaFv3H6mY
Hello @leonvanzyl, How would I save the chat history in flowise? I mean after deploying the model on a website (as shown in your previous videos), I am interested in keeping a record of all the conversations a client has made with the bot and emailed to me along with the lead notification. Is that possible ?
Hey there, you can view the transcripts in Flowise. I believe I did cover it in this video 👍
Great tutorial Leon, just discovered these this weekend. A quick question, I installed flowise on my MacBook and it works great. When I embed it onto a web page, also on my localhost, it works perfectly but there are no avatars / icons. They are there inside flowise but not on the webpage after embedding it. Any suggestions?
Hey, you're welcome 🤗.
Are you referring to the Assistant Avatar? I also noticed that it's not actually used anywhere. I'll ask the Flowise team about this.
Never mind Leon, found it in one of your videos, thank you and a great series
VERY good content. BUT I do get an error "error clearing chat messages" is there any way I can pay for support from you directly?
Try updating Flowise perhaps. Might be a bug.
Thanks for the video! I have a question. May I know the difference between extracting information from PineCone database using index and extracting information from the knowledge files in the OpenAI assistant? I think previously without OpenAI assistant I can also extract data and add custom options to the chatbot?
You're welcome.
The difference is that with OpenAI Assistants, you do not need to use 3rd party services, like Pinecone, to store your knowledge base. It's built into the Assistant API platform.
@@leonvanzyl I tried this with the open ai assistant. If multiple files are loaded, I find that it will only focus on one document and ignore the remaining documents. In other words, it won’t search through all uploaded files in the knowledge files. But using pinecone, all files contents are indexed for search.Do you encounter this problem?
@leonvanzyl I'd really like to learn how to add Speech to text (for user input) and Text to speech (to talk back to the user) to make interaction with the tool way easier. I guess this is custom tooling, do you happen to have a tutorial on this? With openai's whisper and openai's audio speech endpoints this should be possible, however I'm to much of a "no-code" guy to understand how to do it without instructions.
Hi Miles, this sounds like a great idea for an end to end tutorial.
Will see what I can do.
I'm using the endpoint provided by Flowise to integrate the workflow with my own chatbot.
Is it possible to pass a parameter in the body that contains a URL for a doc to upload and to be able to ask it questions?
Leon, what's the difference between a Conversational RAG QA and an Assitant?
Same thing.
A Conversational Retrieval Agent is also known as an AI assistant.
The Assistant in this video is referring to the Assistant API provided by OpenAI. It's an Agent that is fully hosted by OpenAI.
It's still nothing more than a Conversational Agent in the backend.
Use Odoo for my B2B ERP, where the shop is hidden unless logged in. Want to create a Conversational QnA but define a rule to only access that customers data. Where/how would you recommend the rules to be implemented for the postgresql request? Haven’t found any documentation regarding and hopped maybe you had a source document you could recommend, or a video I missed.
Please send my agency an email. It's a bit too much for a comment, but definitely achievable 👍
Awesome tutorial. Just wanted to ask since its uses Assistants API which is paid, what if the user chats about other stuff like hey buy my rayban glasses etc. How do we keep it limited so that it doesnt waste credits?
You could instruct it in the prompt to only answer questions from the knowledge base, and nothing else.
Thanks Leon for another useful video. Can we deploy these assistants to whatsapp, facebook messenger and insta?
Thank you!
I probably wouldn't use Flowise for those integrations at the moment. These apps have incredibly tight timeouts unfortunately, and the assistants API is quite slow on the responses at the moment.
Once this improves, I'll definitely create a tutorial on it.
@@leonvanzyl thank for the reply. What cost effective options do we have to use these assistants on whatsapp, insta and facebook? Voiceflow will be expensive I think
Thank you for sharing. I encountered an issue during testing where the add_lead function doesn't record Chinese characters when the user's dialog text is in Chinese. Is there a solution to this?
would be nice if flowise would expose endpoints for each created agent so we could use them otherwise.
just came around this nice tool. would make agent creation for external use much easier.
What do you mean? Flowise provides API endpoints for each Chatflow.
Great video, I was trying to run it lately but I noticed the Assistants API's latency is pretty bad - sometimes it takes up to a minute for a response to get back. Do you know if there's a way to improve it or is it going to scale up soon?
Thanks
Thank you!
Which model are you using?
I think OpenAI have been battling with the load.
Same one you mentioned over the video
@@nitzantal5365 GPT-4 is definitely a bit slow.
But if I'd like to use assistant functions I must use gpt4, isn't it?
@@nitzantal5365 3.5 Turbo should be able to do it as well and should be a bit faster. Try that
Hi Leon - ive have the chatflow disappear now several times after updating in make. Are there any known issues with flowise ? The assistant still exists in open ai
Did you deploy it to Render? The free plan deleted the flows. Check out my video on deploying Flowise. It shows how you can add persistent storage.
is it possible to stream a openai assistant? also can i connect a exisistng assitant?
Do you have a complete paid course where you can learn all the tools that flowise has because they talk about very few and I know that there are many more that are not found on UA-cam
I'm working on a 2024 Flowise series. What should I include?
I don't see OpenAI Assistants option in the flowise console, what could be the issue? Thanks.
You need to update your Flowise version. Assistants is a new feature.
Hi Leon, thx for this cool Video. I really like the way you describe the entire process making this chatbot with Make and Zoho CRM. But I couldn't make Zoho CRM working, because I do not get a list of Values (connection is valid and verified!) to map the values from the Webhook. If I try with Hubspot, it works. Any Idea, why I do not get the list of possible values for the mapping in the Zoho CRM Module?
You're welcome 🤗.
It's a bit hard to say given the informations, but the ZOHO CRM module should automatically provide all available fields. Perhaps your ZOHO connection isn't set up correctly.
@@leonvanzyl I did setup a second Account on Zoho CRM and also I used a Windows 11 Environment and Edge instead of MacOS Monterey and Safari to test it. I got the same bad result - can't se any fields like Company, Last Name, etc. Not before nor after I established the connection to Zoho CRM as shown in your Video. Are there any API values that one has to setup after creating a Zoho CRM account?
I have the same problem. It only shows me the basic fields, and is missing many of the other ones. The one I need is "Description". If you find a solution or have any alternatives, let me know :D
@@MisterOcean5
I'm trying to connect Hubspot with a "conversational REtrieval QA Chain" but I can't figured out. I'm seeing the only way is with "OpenAI Assistant"? Thank you
What's the error / issue that you're experiencing?
I wouldn't think it would make a difference, since the agent and OpenAI Assistant make use of the same tools and document loaders.
@@leonvanzyl I'm investigating trying to undestand the different types of those chats.
Perfect!!! Just only an input. When I create the under make the zoho crm It is created correctly but the windows where you mach the fields doesn´t appear, so strange.
I think you need to call the webhook from something like Postman first, otherwise Make will have no idea what the possible values looks like
@@leonvanzyl Thanks for answer, I don´t know exactly what you mean and how to do that but thanks.
Thank you a lot man, could we use that with n8n instead of flowise and make knowing that n8n added the option of AI nodes with langchain??
I'm actually looking into creating a completely separate video / series on n8n.
@@leonvanzylthank you man, I think n8n will facilitate more the workflow especially with the recent updates about AI nodes
This guy is a legend! So, so good!
Thank you! This made my day 😄. Just glad I could help
@@leonvanzyl no worries! Keep up the great work 😁
Would love you to make a FW calendly embed option….!!!will see if n8n does this
It's the EXACT same process described in this video. Instead of ZOHO CRM, connect Calendly instead.
Adjust the custom tool and prompt accordingly.
In fact, you could add a second custom tools for getting available dates / times as well.
@@leonvanzyl aha yes makes sense thanks! Will see how calendly looks inside this…Yes I want 2 custom tools…one to check times then send to CRM…this totally rocks…you’ve saved me a ton of time as was on my list for the week!! 👏👏👏👍👍👍
Sir is there any free way to achieve this functionality I mean without using open AI api?
You can use Open source models to achieve some of this functionality, but OpenAI is the leader when it comes to function calling at this stage.
Check my videos on calling Open Source models 😁
@leon I have a question. When creating the custom function, can you also do the same within open ai’s assistant page?
No, you have to "code" the function outside of that interface
@@leonvanzylwhat if you are not using flowise? Can that code still work if added to the assistants interface in functions directly?
The OpenAI interface does not provide any way to enter the code there.
It's a complex and technical process actually. Hence me making this video 😊
Meaning could this still work if you created a json version of it?
Doesn't exist the Actions functionality inside OAI Assistants that could replace the Custom Tool used on Flowise Node? Probably the code will not be same JS based as the example too.
Hi Leon. You are a legend. I have learnt a lot with you! I am loyal subscriber!! Could I ask you someting? I followed your tutorial and it worked with your data. But when I try to write down mine, it gives me this error. Error: Error processing thread: requires_action, Thread ID: thread_T3K40MVIeeAzu0ET7OUAVPKp, Run ID: run_Lzs714A7v078l2pzc5FmiyKB, Tool: add_lead
It is exactly the same as you did, I just added two more functions besides name, phone and some others. Do you know why this is? I have been trying the whole day to fix it and it is really driving me crazy :(
Thank you for the compliment 🤗.
It's really hard to say what the issue could be based on only that, but maybe someone in the comments can assist.
@@leonvanzyl I could fix it. It's taken me more than 6 hours to figure it out, but it's alright now. Thanks for the response anyway! Looking forward to your next videos!!
Hi. Is it possible to use this on a Shopify website ?
Yes! You just need to be able to add script tags to the theme.
This might be a good tutorial idea.
Thanks Leon, great tutorial!
Thank you!
The template works wonderfully. However, my bot disappears in the chat flow of Flowise after about 2 hours and is not permanently saved. What could be the reason for this?
The free account does not have permanent storage. Check out my deployment video to see how to set up a paid account with storage.
If you're using this for business or clients, the $7 p/m is well worth it.
@@leonvanzyl Thank you! That explains the problem.
Thank you for this video!
You're welcome!
I have two questions.
Is there any difference between assistent and custom gpts ?
and How can i integrate them in instagram or whatsapp ?
Custom GPTs are accessed by the ChatGPT website and are basically custom versions of chatGPT.
The Assistants API is not related to Custom GPTs and does not have the same behaviour or functionality. Instead, it a framework that developers can use to build advanced agents / chatbots.
As for WhatsApp/ Intragram integration, it's way too complex to answer in a comment. In a nutshell, you need to familiarise yourself with each of those platforms' developer tools and how their webhooks work.
@@leonvanzyl but is it possible to integrate whatsapp instagramm ?
you deserve more subs and views
Wow, thank you!
Appreciate it.
Please like my videos. It helps the algorithm to show my videos to more people 🙏
Is it possible to use an Agent that can call tools just like in this example but also perform lookups on a vector database? Kinda like a mix of Conversational Retrieval QA Chain and OpenAI Assistant.
Yes, that is exactly what the assistant does. Simply upload document to its knowledge base and the agent will automatically figure out when to use the knowledge base (a vector store) or call tools.
@@leonvanzyl But that's quite limiting compared to a classic rag with a vector database where I can store vector representations of any kind (and amount) of data, be it code or text split in the way that I'd consider best suitable for it (chunk sizes and so on).
@@RobertoFabrizi yeah, that's one of the disadvantages of OpenAI Assistants.
Building your own solution gives you way more control and you are not locked into the vendor.
Trade-offs 😁.
@@leonvanzyl If I may ask one last question on the matter (thank you so much for your time!!) how would you suggest that such use case be implemented? Leveraging the ReAct agent to invoke che custom tool?
@@RobertoFabrizi I am working on an updated series on Flowise where I cover exactly this. Stay tuned 😁. It's way too complex to explain in a comment.
When I embed it, I can not scroll in the desktop version of the chatbot. Is there a fix?
For interest sake, what do you mean by "desktop version"? Are you referring to the full screen embed on a website?
@@leonvanzyl No, I mean the Pop-up chatbot on my wordpress website. The problem is that in the mobile version I can scroll through the messages with my finger but in the desktop version I have no way to scroll up or down in the chatbot. Even if i hover over the chatbot with the mouse pointer, the entire website scrolls but not the chatbot
Hi Leon, learning so much from you, thank you. Are you available as a freelancer to help me with some finer details?
You're welcome 🤗.
You can send a message to my agency and I'll see if it's something I can assist with. Check the description.
@@leonvanzyl done, thank you. (I am imagining your agency to be a complex network of AI bots! haha)
@@leonvanzyl Hi Leon, haven't heard anything back from you so I'm not sure if my enquiry came through. Do you mind checking? My name is Cam L.
@@leonvanzyl hi mate, haven't heard back from anybody. Do you mind checking if you've received my request?
Hi leon!
thanks for sharing such an amazing content with us... I was trying to integrate a flowise powered chatbot on my worpress website and followed all the steps mentioned in the video but getting error... the error is about rule 941110 - reason XSS Filter... can you please help in resolving it, I ll really appreciate it...
Haven't seen that error before. Reach out to me via email and we can schedule a call.
@@leonvanzyl ok thanks
Don't you just hate it? Following tutorial step by step and when it gets to testing it's not working. Set up the assistant, uploaded the file with company info and when asking questions (which are included in the document) it tells me that it couldn't`t find the answer in the document provided. Checking with Open AI looks like the document is uploaded and tokens getting added like it is working but I get only negative answers.
What happens when you test it in die OpenAI playground? Same issue?
@@leonvanzyl In playground mode looks like it's working.
I'm struggling with this for 3 days now> I redone the assistant 4 or 5 times, deleting all previous and starting from scratch
@@leonvanzyl In Open Ai/Assistant at the bottom of the page is a field called API Version with an option to switch to v1. If we switch to v1 will it have an effect?
@@leonvanzyl This is useless man. I deleted everything and started from scratch 7 times. Watched your video 20 times just in case I missed something. It is not working. In Playground I get the right answers but in Flowise tells me it can not find the specific information in the document provided and I asked the same question. Is there any other way to do this?
Amazing video, thank you
Thank you
Brilliant. Thanks Leon
You're welcome 🤗
I noticed when the information is retreived it contains its source, like 【9†source】. How do you clean them up from the output?
Noticed that too sometimes.
I've reached out to Flowise so they can hopefully address it.
The Assistants API natively returns the citations on the responses, and it's up to the consumer (Flowise in this instance) to remove the citations.
Hopefully OpenAI will parameterize that functionality soon.
@leonvanzyl I've noticed that Flowise has some output parsers, but none of them allow me to run custom JavaScript code to clear those entities out. If you receive feedback from them, please let me know. I appreciate it! ;)
@@RussianBritish just got feedback from Flowise. Their pushing an update SOON to remove citations.
@@leonvanzyl incredible news Leon. Can't wait to see it! Then I will move away my chatbot from Voiceflow to Flowise
thank you so much , this just literarily save my life
You're welcome 🤗
just wonder, after crating the assistant on Flowise, can I add functions through the Openai website? or stll have to add through flowise? it would be absolutely great if we could do modification later directly on their openai page
Thank you very much, great Video.
You're welcome 🤗
I can combine RAG chat bot and Assistant into one chat flow?
Keep in mind that the Assistant is a RAG chatbot. I assume you're asking if you can add more than one agent to the same chatflows.
I suppose it is possible, but it will involve calling different chatflows using they're APIs.
@@leonvanzyl i make a chat bot which speak on my own information (like in u RAG tutorial)
And now i want to add booking function into this bot.
And i have second question, how i can make the text bold in my bots response (like on u website)
How is it possible, that I don't have the node openai assistant?
Update your version of Flowise
@@leonvanzyl We are using 1.4.3 and do not see the Open AI Assistant under agents. Is there some trick to get this working? Thank you for your amazing videos. They have been so helpful.
thank you for your video, but I don't know why I make step by step like you do, but at 11:00 when I chat Hello, it show an errror Message like this: "Error: Error processing thread: failed, Thread ID: thread_mdI12dqWexCGUXv6d1ZXTHVB, Run ID: run_fmjL58EGTobM06y6IGbE0orB, Status: failed". Can you help me to resolve the problem. Thank you so much !!!
You're welcome 🤗.
Clear the chat and try again. Also upgrade Flowise and try again.
WoW really nice video thank you so much Leon 🙏🙏🙏💚💚💚👏👏👏
Thanks Trax, glad you enjoyed it 😊
Amazing. Thank you 🙌🙌
Amazing comment, thank YOU 😁
Excellent video tutorial! 👏
Excellent comment! Thank you