You just made my day!!! Legend!! Spent a week trying to get this sort of working in JS and make..Calendly is a nightmare as they don’t have a true scheduling API…😀😀😀
@@leonvanzyl you deserve it! er quick question - im seeing really poor return times back to openai/FW with the booking link. Make is completing the request very fast end to end in under a second but taking forever to get the link. Have upgraded my DO server to 2gb. Any thoughts? Is it an openai beta thing? Also can you code the URL link structure inside the JSON for this function at all? Thanks
Hi Leon, how are you? As always you surprise me with something useful and practical!! Please keep it that way 😊 . One simple question. I've seen you use a lot of OpenAi Assistants, but I havent seen you enter openai apii keys. Do openai assistants incur in openai tokens spent? Thank you!!
Great video Lyon, I have two questions please: 1. Does this function aligned with openai GPTs' actions? Meaning if I copy/paste that code and schema to openai GPTs, should it work? 2. Is there an option to authenticate a user to oauth provider (e.g. Google) by responding an authentication link same as you did for calendy, but once the user had been authenticated, may I chain his auth token to another function (sort of function chaining which involves user interaction)? Thanks, you're the best
Thank you for the feedback 🙏. I believe point 1 will work. It's all based on OpenAI function calling after all. I doubt point two will be possible though, or will be interesting to see how it would be implemented. Token authentication (and renewal) is quite a complex process on the server and I'm not sure how Flowise could handle it. I'll chew on it a bit more.. 😎
Thanks@@leonvanzyl Maybe I didn't explain myself clearly. The flow is as is: function_1 returns Google auth link -> user clicks the link -> auth window pops up -> user approves authentication -> auth accepted -> pop us closed -> function 2 runs with auth token. I hope it's more clear now, thanks
@nitz that won't be possible as human interaction within a tool is not possible. When a tool or Custom Action is invoked it will without human intervention
Hey can we provide the available date and time ? Right there in the chatbot and collect all the necessary information required to book an appointment , ??
@@leonvanzyl thank you for replying , how can I do that ? Please make a video on that , so that we can we book an appointment right there in the chatbot
Thank you very much for your videos with the detailed explanations. They have helped me a lot! As I want to keep my data locally, I would like to see a video about this. So local LLM's (I use LM-Studio), local embeddings, local database, local vectorstore etc.. It would also be interesting to send data to Excel or via XML to an ERP system.
@leonvanzyl first, thank you very much for your videos! I have a question about tools. I just created a custom tool, that would show up a video in the chatbot or an image. in the playground of flowise it functions but on the website embedded, it maybe sanitized the code to showup a video to be a string and so it shows up the full -code in the chatbot-window. is there a solution for this?
Love the work as usual! Anyone reading this comment definitely needs to go and check out the other videos! A quick question, can you train the answers the chatbot gives? I’m not a huge fan of some of the answers it gives, especially with the latest update and new llm open ai node.
Awesome man! As a educationalist and instructional designer I made a doc searcher with flowise and assistent. This is also very cool. Do you have an idea on how to change the 'welcome' massage at the start. Also, can you combine flowise with voiceflow?
That sounds awesome! I'll love to see your project in action. You can change the welcome message in the embed config (see my web embed video). Yes, it's easy to call Flowise from Voiceflow. Simply add an API node in your Voiceflow flow and call the Flowise API endpoint..I have a video on Flowise API endpoints as well 😁
@@leonvanzyl you are the best man! Thanks for all the work for the community and people from other disciplines learning to deal with AI. Looking forward for the next video! My project is realy easy tbh. Just a retriever tool as allowed tools for a assistent api. But the cool thing is that the assistent is using it as a tool in OpenAi dashboard as well! Then the AI search the knowledge base for the provide pdf (course material). I am now testing if it can search multiple docs and still can find the right answers. Our students are learning precise legal texts, so there is little margin for error for the chatbot. It does a really nice job tbh, even on the Dutch language. Based on your last name, do you have dutch/germanic heritage as well? Edit: I now know how to change the welcome message. Did not saw the "show embed chat config" button.
Why go to all those steps when you can just set this in the custom tool: return 'YOUR CALENDLY LINK' and skip the whole Make connection? cause that is what you are doing with a lot of extra steps right or am I missing something 🙂?
You could definitely just return a static link if you wanted. You can use these integrations to do a lot more though, like generate once off links and dynamically change parameters based on human input. Either way is fine.
I was wondering the same. It would be great to first capture a lead in a CRM like Zoho, and after that asking the user if they want to schedule an appointment themselves (using the lead information that is already provided)
Is it possible to book an appointment directly from the chatbot , without redirecting to the link ? , I mean can we get all the necessary information from the user and book an appointment right there in the chatbot ? Is it possible?
hi! is there a possibility to integrate a logic, like if the user has more then 5 questions, the chatbot will say something like: "I see that you have many questions. Would you like to have an appointment with us? Leave your name, email and phonnumber..." this will prevent users chatting long with the chatbot and inducing it to get leads
@@leonvanzyl thanks for your answer. I tryed that but it doesnt allways work. I have worked with voiceflow and there you could put such a logic. I put there a counter that counts +1 after every question and after the fifth question it triggers the question for getting a lead. I am wondering if it is possible something like that in flowise.
Amazing tutorial! I was wondering if it would be possible to combine a Conversational Retrieval QA Chain with an AI Agent like this to not only book appointments with a tool but also use a pinecone to retrieve information to answer questions. I tried to make them but combining them to be one AI proved to be confusing. Is there a way? Thanks!
Thank you! Yes, you can combine agents with chains AND vector stores using tools. Check out this video on how to add a retriever and chain as tools to an agent. ua-cam.com/video/Dinzg_dGfqU/v-deo.html
Hi león Great video. Have achieved this before, however been trying to make it more natural by talking with the assistant to book an appointment. example: I would like like to book a call on Thursday at 9am. I have had no luck, is it not possible to call calendly's api and add this to a function to the assistant?
Calendly does not allow for this functionality unfortunately. They only offer a scheduling link solution. I think the user experience would be quite bad if you had to have a back and forth with a bot to find an open date and time, AND somehow process payments in order to confirm the booking 😁.
Ah ok. To make it a better experience it could go something like this: I would like to book an appointment > what day would you prefer > Friday > Friday we have 3pm and 5pm available. Im wondering if there is any other way to achieve this @@leonvanzyl
@nzyl Thanks to your videos I was able to have a custom tool using conversation retrieval agents or conversation agents and a retrieval tool querying the vector database on supabase. Awesome work bro
Question: if the goal is to get the assistant answer with the calendly link... It's not easier to just give to the assistant the calendly link on the instructions? This way we don't need make / webhooks and achieve very similar result. Or am I missing something?
These are single use links, and you could enhance this process to dynamically set the event types based on the user's questions. But yes, if you really wanted to, you could just tell it in the instructions to provide a static Calendly link for a specic event type 👍.
You just made my day!!! Legend!! Spent a week trying to get this sort of working in JS and make..Calendly is a nightmare as they don’t have a true scheduling API…😀😀😀
This comment made my day! Thank YOU 🙏
@@leonvanzyl you deserve it! er quick question - im seeing really poor return times back to openai/FW with the booking link. Make is completing the request very fast end to end in under a second but taking forever to get the link. Have upgraded my DO server to 2gb.
Any thoughts? Is it an openai beta thing?
Also can you code the URL link structure inside the JSON for this function at all? Thanks
Thank you for creating such a great content about flowise , it really helped , make more videos like this
First! Another excellent tutorial Leon! Just what I was waiting for. You have the best Flowise content.
Thank you Alton.
Simply awesome!!
Thank you for sharing this, just what I was searching for these past few days...👍
Glad to hear 😊. You're welcome
Thank you for your awesome content !!!!!! I´m really really thankful !!!!
You're welcome 🤗
Hi Leon, how are you? As always you surprise me with something useful and practical!! Please keep it that way 😊 . One simple question. I've seen you use a lot of OpenAi Assistants, but I havent seen you enter openai apii keys. Do openai assistants incur in openai tokens spent? Thank you!!
Hi Ricardo,
Thank you for the feedback!
The openai assistants do require an API key. I might have set it during the creation of the Assistant
@@leonvanzyl Good to know!! Thank you!!
Are there any alternatives that you know about to replace the OpenAI assistant in this use case?
Great video Lyon,
I have two questions please:
1. Does this function aligned with openai GPTs' actions? Meaning if I copy/paste that code and schema to openai GPTs, should it work?
2. Is there an option to authenticate a user to oauth provider (e.g. Google) by responding an authentication link same as you did for calendy, but once the user had been authenticated, may I chain his auth token to another function (sort of function chaining which involves user interaction)?
Thanks, you're the best
Thank you for the feedback 🙏.
I believe point 1 will work. It's all based on OpenAI function calling after all.
I doubt point two will be possible though, or will be interesting to see how it would be implemented. Token authentication (and renewal) is quite a complex process on the server and I'm not sure how Flowise could handle it. I'll chew on it a bit more.. 😎
Thanks@@leonvanzyl
Maybe I didn't explain myself clearly. The flow is as is:
function_1 returns Google auth link -> user clicks the link -> auth window pops up -> user approves authentication -> auth accepted -> pop us closed -> function 2 runs with auth token.
I hope it's more clear now, thanks
@nitz that won't be possible as human interaction within a tool is not possible. When a tool or Custom Action is invoked it will without human intervention
Hey can we provide the available date and time ? Right there in the chatbot and collect all the necessary information required to book an appointment , ??
You could create a custom tool for retrieving the current date and time ☺️
@@leonvanzyl thank you for replying , how can I do that ? Please make a video on that , so that we can we book an appointment right there in the chatbot
And why don't you include the Calendly reservation link in the same prompt without using the Make tool?
Same question here 🙏
Your all videos are very interesting and useful. Can you create video regarding streaming effect in UI? This will be very useful.
Thank you very much for your videos with the detailed explanations. They have helped me a lot!
As I want to keep my data locally, I would like to see a video about this. So local LLM's (I use LM-Studio), local embeddings, local database, local vectorstore etc..
It would also be interesting to send data to Excel or via XML to an ERP system.
Thank you for the feedback 🙏
@leonvanzyl first, thank you very much for your videos! I have a question about tools. I just created a custom tool, that would show up a video in the chatbot or an image. in the playground of flowise it functions but on the website embedded, it maybe sanitized the code to showup a video to be a string and so it shows up the full -code in the chatbot-window. is there a solution for this?
Love the work as usual! Anyone reading this comment definitely needs to go and check out the other videos!
A quick question, can you train the answers the chatbot gives? I’m not a huge fan of some of the answers it gives, especially with the latest update and new llm open ai node.
Thank you for the kind feedback.
You can use the Assistant Instructions to change its behaviour and tone. Maybe you could try that?
Just a question, can't you just ask Flowwise to display the booking URL, without going through all that steps?
same question
Yes!
Awesome man! As a educationalist and instructional designer I made a doc searcher with flowise and assistent. This is also very cool. Do you have an idea on how to change the 'welcome' massage at the start. Also, can you combine flowise with voiceflow?
That sounds awesome! I'll love to see your project in action.
You can change the welcome message in the embed config (see my web embed video).
Yes, it's easy to call Flowise from Voiceflow. Simply add an API node in your Voiceflow flow and call the Flowise API endpoint..I have a video on Flowise API endpoints as well 😁
@@leonvanzyl you are the best man! Thanks for all the work for the community and people from other disciplines learning to deal with AI. Looking forward for the next video!
My project is realy easy tbh. Just a retriever tool as allowed tools for a assistent api. But the cool thing is that the assistent is using it as a tool in OpenAi dashboard as well! Then the AI search the knowledge base for the provide pdf (course material). I am now testing if it can search multiple docs and still can find the right answers. Our students are learning precise legal texts, so there is little margin for error for the chatbot. It does a really nice job tbh, even on the Dutch language.
Based on your last name, do you have dutch/germanic heritage as well?
Edit: I now know how to change the welcome message. Did not saw the "show embed chat config" button.
Why go to all those steps when you can just set this in the custom tool: return 'YOUR CALENDLY LINK' and skip the whole Make connection? cause that is what you are doing with a lot of extra steps right or am I missing something 🙂?
You could definitely just return a static link if you wanted. You can use these integrations to do a lot more though, like generate once off links and dynamically change parameters based on human input. Either way is fine.
I was wondering the same. It would be great to first capture a lead in a CRM like Zoho, and after that asking the user if they want to schedule an appointment themselves (using the lead information that is already provided)
Is it possible to book an appointment directly from the chatbot , without redirecting to the link ? , I mean can we get all the necessary information from the user and book an appointment right there in the chatbot ? Is it possible?
hi! is there a possibility to integrate a logic, like if the user has more then 5 questions, the chatbot will say something like: "I see that you have many questions. Would you like to have an appointment with us? Leave your name, email and phonnumber..."
this will prevent users chatting long with the chatbot and inducing it to get leads
You could just add that instruction to the system message. The agent should be able up handle that
@@leonvanzyl thanks for your answer. I tryed that but it doesnt allways work. I have worked with voiceflow and there you could put such a logic. I put there a counter that counts +1 after every question and after the fifth question it triggers the question for getting a lead. I am wondering if it is possible something like that in flowise.
Amazing tutorial! I was wondering if it would be possible to combine a Conversational Retrieval QA Chain with an AI Agent like this to not only book appointments with a tool but also use a pinecone to retrieve information to answer questions. I tried to make them but combining them to be one AI proved to be confusing. Is there a way? Thanks!
Thank you!
Yes, you can combine agents with chains AND vector stores using tools.
Check out this video on how to add a retriever and chain as tools to an agent.
ua-cam.com/video/Dinzg_dGfqU/v-deo.html
@@leonvanzyl Thank you! Exactly what I looking to learn :)
@@MyXRLearning you're welcome 🤗
over-enginered. You could just send the calendly link as a response in the chatbot right? or i am missing something here?
Definitely an option
Another gem from you. Thank you
Thank you 🙏
Great tutorial - Thank you
Thank you 😊
Excellent video!👍
Thank you 😊
Flowise and OpenAI WILL be great but I got a lot of error "Thread..." or the bot does the action twice. Do you have these kind of issue ? Thank you
Haven't had any issues like that. Ensure that your Flowise version is up to date?
Does flowise intend to make you use webhook? am i reliant on that, or are there other alternatives?
You can call any API endpoint from tools 👍
that's great content. thanks Leon
Thank you!
Amazing , i try this also in n8n work nicely
That's awesome!
Are you self hosting n8n?
@@leonvanzyl yes , on railway
Hi león Great video. Have achieved this before, however been trying to make it more natural by talking with the assistant to book an appointment. example: I would like like to book a call on Thursday at 9am. I have had no luck, is it not possible to call calendly's api and add this to a function to the assistant?
Calendly does not allow for this functionality unfortunately. They only offer a scheduling link solution. I think the user experience would be quite bad if you had to have a back and forth with a bot to find an open date and time, AND somehow process payments in order to confirm the booking 😁.
Ah ok. To make it a better experience it could go something like this: I would like to book an appointment > what day would you prefer > Friday > Friday we have 3pm and 5pm available. Im wondering if there is any other way to achieve this @@leonvanzyl
@@tas5247 definitely. I'll add more videos to this series.
@@leonvanzyl look forward to it!
you are the TRUE GOAT.
Hahaha! This is a first. Thanks 😁
Can we do the same with Conversational Retrieval QA Chain with Supabase vector database instead of Openai assistant? Thanks
I am releasing a video in creating RAG agents (aka Assistants) in the next day or two.
@nzyl Thanks to your videos I was able to have a custom tool using conversation retrieval agents or conversation agents and a retrieval tool querying the vector database on supabase. Awesome work bro
Question: if the goal is to get the assistant answer with the calendly link... It's not easier to just give to the assistant the calendly link on the instructions? This way we don't need make / webhooks and achieve very similar result.
Or am I missing something?
These are single use links, and you could enhance this process to dynamically set the event types based on the user's questions.
But yes, if you really wanted to, you could just tell it in the instructions to provide a static Calendly link for a specic event type 👍.