sure thing, working on trying more enterprise level stuff, will start making my content slowly more tech advanced so make sure to hit that notification bell 😉
I would probably use this for a lot more advanced AI memory. Because if you are using it to pass up memory and provide incredible context to the model, you use a 'SET' inside of Voiceflow and have a "No match" inside of your intents card. So if the "last_utterance" did not match any other intents inside of your "Intents" card you can just loop it back to the "Node" that has AI model with "knowledge base" turned on. And that set step will be the one that actually handles the memory and context. I haven't put it through a real test yet but the "SET" card can handle quite a large amount of context because it just accumulates the "User's" questions and what the AI responded with. I'd probably use this step to cut cost and make my model answer questions like, "how heavy is John Cena?"🤣. But great tutorial by the way.
not a bad idea, I think this is definitely a work around to lower the consumption but given that you can host your own LLMs on your computer just using flowise to pass it your own AI is probably superior choice. When I made this tutorial this was definitely not easily accessible for the common folk but now I can run fine-tuned Mistral Model on my Mac and have external IP listening for requests from Voiceflow bot and provide inference 2 times faster than Openai or Anthropic :) I will be making that video soon
Do you know of any tutorials or documentation showing how to integrate with Protopie or Framer to build and launch a app UI? Thanks for the video brother!
Awesome video, just subscribed ! One question : Can flowise or other local open source that can do this kind of chat diagram ? (mutliple choice chat and guide the discussion) I used flowise to build open chatting, but I'm not sure how to guide it with questions and answers. Thanks, you rock !
Thanks man. Will try to get few useful videos out this week. I think there is a big void in creating custom knowledge bases and why they are better than what tools like Botpress and Voiceflow provide
Great. I have a question. When you ask follow-up questions I see that you have to pass the history: Client is asking follow up question: {last_response} Chat History: Client- {portable restroom ASIMP}, AI Answer {flowise_answer1} How do you recreate the history dynamically at each iteration?
This is AMAZING, my friend! You are definitely taking things to the next level. I am jumping into your Discord as we speak. When will you be releasing your course?
Hi, AI Simp! Well, I want to develop a chatbot mobile app in flutterflow, developed in botpress, if not possible even voiceflow/flowise. Is this possible? If it is I think a tutorial about this can be of great use to the community, thanks!
If you go flutterflow route you don’t need Botpress, just attach flowise and you are golden. Botpress is just a nice GUI, flutterflow is also a nice gui. I will add intro classes for flutterflow to my course. I build chatbots and flutterflow apps. It is a natural progression of learning. Glad you saw it too.
@@aisimp So will you show in your course how to create a custom chat app? I want to add a custom chat to a FF app, but in botpress I see more customization into the chatflow. But also if you can show how to directly do it in FF would be amazing, thanks a lot 🙏
Answer to my prayers man :) And the bot can be deployed on any channel i.e. SMS or Whatsapp this way? Ive beem mostly building them in HL and Zapier, using SMS as main channel but really curious how to make it work multi-channel seamlessly and ideally, continue supporting the conversation with the user on a different channel, as long as some unique ID like a phone number (fro SMS and Whatsapp), is the same.
I think it is possible, Flowise is the server and how you are connecting to it can be customized based on the use case. Shameless plug sign up for the course , we will be trying advanced use cases that are not on the UA-cam 😉
Can i try your chat bot built with Voiceflow & Flowise? Or any alternative support voice input & output (aka: voice assistant). Thank so much for your video.
Fantastic combination of new AI tools! Will definitely help with my work and also look out for your new course. Thanks! BTW, the latter part of the video made me a bit😵💫
Absolutely love the detailed explanation in your videos. Thank you! Quick question: am I right that when you send a request to Flowise, you don't pass the history, and so it would not work well with follow-up questions? Is there a way to overcome this limitation?
So in my use case, flowise is keeping historical log of messages, if that wouldn’t be the case I would add history from Voiceflow so my LLM model had a context and provided a better answer.
a quick question tho, can you prompt instructions to Flowise before processing the question from clients? would this step be in voiceflow or flowise? thanks a lot!
@@aisimp i'm not asking for an easy way my friend. i'm asking what you did (coded or created a script to upsert the docs.) based on your answer i'm gonna have to create a script, lol
Hey mate! Great video!! I have done everything and throws at me the code 200. But when it comes to give me the answer, it was objetc, object. API CAll successfully triggered but it gives me that answer. [Object Object]. What could it be? Some help would be really appreciate it! Just subscribed to your channel ;) Keep up the good work!
I don't understand something, why pass by flowise for a chatbot? Voiceflow already do the job perfectly right. Why bother with 2 different layers of cvhatobt, while you can only use one
Yeah I was wondering this too? Since the Voiceflow knowledgebase is super easy to update/manager and also uses it's own vector database with seemingly excellent responses. I guess if we were to add a web browser or an agent in Flowise then that's something Voiceflow can't do.
I am guessing, for flexibility with agents, segregating into multiple knowledge bases for different responses within different phases of a large chat sequence, and the cost of token being cheaper.
Also if your company wants fine-tuned LLM model such as Llama2 or Falcon, results are better than using Openai and many companies don’t feel very jolly about sending their corporate info to Openai
Thanks you for this epic tutorial! Question: Have you noticed big speed differences when querying Pinecone vs. Quadrant (in terms of chatbot response times)? I'm currently all set up on Pinecone so migrating isn't ideal but if Quadrant is way faster, I'd for sure make the move. Also, I'm hoping for Voiceflow to stream it's chatbot responses (to improve the UX), but I'm guessing that's not possible when we make calls to Flowise? Appreciate it again! New subscriber over here
I would say they are probably very close, probably what matters more is the rest of your stack, here a difference between them probably 0.1 seconds so not a big deal. Well my workaround with streaming is just try to limit the length of responses this way the answer come faster, so instead of one step with long response truncate it in two smaller steps to avoid long lag. Also if you decide that you need streaming you can always just go full code and build it on NextJS
@@grindwithmal no worries, feel free to join the discord I am making a collection of the best tech/tutorials in the ai space for developers, I am not biased 😂 I mention other UA-camrs/bloggers if they are good
@aisimp Hi there! Great content dude. Some quick questions - Voiceflow has knowledge base so when does it make the case to use Flowise / Stack-ai? I want to make a training bot which references documents and URLs in a sharepoint - that is not scaleable on Voiceflow so what tech stack would you use and why? If I wanted to reduce the cost of tokens whats a good LLM to use whilist maintaining accuracy? Thanks in advance EDIT: Just saw previous post had similar question but if anything anyone can add. VF alone can do a decent job but not scaleable I take it If I want to crawl 1000s of urls?
@@aisimp I have gotten it to run on Render, but after a few hours when I come back and reload the page, my project is no longer there. Have you experienced this?
I am filming right now but bunch of ideas that I will have in the course I share in my AISIMP dev letter. In the future I can just reference each letter (they are numbered) to help with understanding.
Funny that you ask, I connected Telegram bot to whisper to STT and used elevenlabs for STT. There is a slight delay because elevenlabs can’t export in OGG format so I format it from mp3 to OGG and then send it to Telegram, all that done in Flowise and Telegram is just the front instead of Voiceflow
@@musumo1908 sounds good, I think I will be accumulating all of these prototypes somewhere cause editing these videos takes forever and the video editor is a secret 🤫 … no one is sharing it, found it on twitter
Dude you already are pretty well worse in flowise Ai and stack Ai is also mostly same its really confusing which one to chose. Can you make a video on it
Voiceflow is using Openai for AI functionality and then charges you 2.5x more cost wise vs just using Openai directly and Flowise is free. So here is the reason for that statement
Not really.m Voiceflow fee is calculated based on the usage of Ai Tokens, but if you run out of your plan then you need to upgrade or buy more at that extra price. People pay because voiceflow appears easier to use, other than that it is on parity with many other applications@@RolandoLopezNieto
@@aisimp I don't quite understand those API integrations, intents, entities, those variables and so on. So I request you to create a simple one with both of them being used.
Gold, pure GOLD! Thank you man for the amazing content! 🙏🏼
sure thing, working on trying more enterprise level stuff, will start making my content slowly more tech advanced so make sure to hit that notification bell 😉
@@aisimp
Already did! 😁
I would probably use this for a lot more advanced AI memory. Because if you are using it to pass up memory and provide incredible context to the model, you use a 'SET' inside of Voiceflow and have a "No match" inside of your intents card.
So if the "last_utterance" did not match any other intents inside of your "Intents" card you can just loop it back to the "Node" that has AI model with "knowledge base" turned on.
And that set step will be the one that actually handles the memory and context. I haven't put it through a real test yet but the "SET" card can handle quite a large amount of context because it just accumulates the "User's" questions and what the AI responded with.
I'd probably use this step to cut cost and make my model answer questions like, "how heavy is John Cena?"🤣.
But great tutorial by the way.
not a bad idea, I think this is definitely a work around to lower the consumption but given that you can host your own LLMs on your computer just using flowise to pass it your own AI is probably superior choice. When I made this tutorial this was definitely not easily accessible for the common folk but now I can run fine-tuned Mistral Model on my Mac and have external IP listening for requests from Voiceflow bot and provide inference 2 times faster than Openai or Anthropic :) I will be making that video soon
Do you know of any tutorials or documentation showing how to integrate with Protopie or Framer to build and launch a app UI? Thanks for the video brother!
exactly what I needed, you are a treasure man! huge thanks!
Sure thing, also forgot to mention that you can do most of the functions people pay for Stackai for only $7 using flowise 🫡
Absolutely love this!
Thanks 🙏🏻, congrats on the new Funding Round. Would love to learn more what you guys have in store as far as AI AGENTS
Very nice tour on what you can achieve. Where can I check out the course you mentioned? Thank you!
Awesome video, just subscribed ! One question : Can flowise or other local open source that can do this kind of chat diagram ? (mutliple choice chat and guide the discussion)
I used flowise to build open chatting, but I'm not sure how to guide it with questions and answers.
Thanks, you rock !
Thanks for sharing your progress, look forward to the follow up videos, subscribed
Thanks man. Will try to get few useful videos out this week. I think there is a big void in creating custom knowledge bases and why they are better than what tools like Botpress and Voiceflow provide
Great. I have a question. When you ask follow-up questions I see that you have to pass the history:
Client is asking follow up question:
{last_response} Chat History: Client- {portable restroom ASIMP}, AI Answer {flowise_answer1}
How do you recreate the history dynamically at each iteration?
This is amazing, saving me a lot of money and better delivery for my clients 🔥
is using Flowise better than using Response AI on voiceflow? could you tell me the benefit of using Flowise?
I never used Response AI. Why do you like it?
Priceless. Thank you so much
Sure my pleasure 😇
This is AMAZING, my friend! You are definitely taking things to the next level. I am jumping into your Discord as we speak. When will you be releasing your course?
Still building, need another week
Hi! When should one opt to fine tune instead of using a Voiceflow / Botpress bot for Q&A ?
Hi, AI Simp! Well, I want to develop a chatbot mobile app in flutterflow, developed in botpress, if not possible even voiceflow/flowise. Is this possible? If it is I think a tutorial about this can be of great use to the community, thanks!
If you go flutterflow route you don’t need Botpress, just attach flowise and you are golden. Botpress is just a nice GUI, flutterflow is also a nice gui. I will add intro classes for flutterflow to my course. I build chatbots and flutterflow apps. It is a natural progression of learning. Glad you saw it too.
@@aisimp So will you show in your course how to create a custom chat app? I want to add a custom chat to a FF app, but in botpress I see more customization into the chatflow. But also if you can show how to directly do it in FF would be amazing, thanks a lot 🙏
Nice! When will you launch your course?@@aisimp
Answer to my prayers man :) And the bot can be deployed on any channel i.e. SMS or Whatsapp this way? Ive beem mostly building them in HL and Zapier, using SMS as main channel but really curious how to make it work multi-channel seamlessly and ideally, continue supporting the conversation with the user on a different channel, as long as some unique ID like a phone number (fro SMS and Whatsapp), is the same.
I think it is possible, Flowise is the server and how you are connecting to it can be customized based on the use case. Shameless plug sign up for the course , we will be trying advanced use cases that are not on the UA-cam 😉
Can i try your chat bot built with Voiceflow & Flowise? Or any alternative support voice input & output (aka: voice assistant). Thank so much for your video.
Fantastic combination of new AI tools! Will definitely help with my work and also look out for your new course. Thanks! BTW, the latter part of the video made me a bit😵💫
Glad you liked it. Just released www.thetacombinator.ai/yt-presale-23
Absolutely love the detailed explanation in your videos. Thank you! Quick question: am I right that when you send a request to Flowise, you don't pass the history, and so it would not work well with follow-up questions? Is there a way to overcome this limitation?
So in my use case, flowise is keeping historical log of messages, if that wouldn’t be the case I would add history from Voiceflow so my LLM model had a context and provided a better answer.
@@aisimp Great vid just subscribed how do you add history from Voiceflow though?
a quick question tho, can you prompt instructions to Flowise before processing the question from clients? would this step be in voiceflow or flowise? thanks a lot!
Yes I have prompt in the flowise so when voiceflow queries it, voiceflow prompt is just a human entry, the prompt for the function is in flowise
Great content! Did you create another flow to upsert documents into qdrant collections? how did you create the collections in the first place?
Yeah okay how about I will show you a super easy way to get vectors up in Qdrant super fast in 2 mins or less
@@aisimp i'm not asking for an easy way my friend. i'm asking what you did (coded or created a script to upsert the docs.) based on your answer i'm gonna have to create a script, lol
@@pedromussi you just pass the files to the upstert module via REST API
Nice! What are you using to create the videos?
Secret go to my discord I will tell you there 😅
Hey mate! Great video!! I have done everything and throws at me the code 200. But when it comes to give me the answer, it was objetc, object. API CAll successfully triggered but it gives me that answer. [Object Object]. What could it be? Some help would be really appreciate it! Just subscribed to your channel ;)
Keep up the good work!
Thanks for the insight.This was a big question on mine that you answered.
I don't understand something,
why pass by flowise for a chatbot?
Voiceflow already do the job perfectly right.
Why bother with 2 different layers of cvhatobt, while you can only use one
Yeah I was wondering this too? Since the Voiceflow knowledgebase is super easy to update/manager and also uses it's own vector database with seemingly excellent responses.
I guess if we were to add a web browser or an agent in Flowise then that's something Voiceflow can't do.
I am guessing, for flexibility with agents, segregating into multiple knowledge bases for different responses within different phases of a large chat sequence, and the cost of token being cheaper.
@@titusmulquiney8124 ye was thinking the same.
What do you think would be a key bonus feature for a lead generation bot?
Also if your company wants fine-tuned LLM model such as Llama2 or Falcon, results are better than using Openai and many companies don’t feel very jolly about sending their corporate info to Openai
Exactly
Thanks you for this epic tutorial! Question: Have you noticed big speed differences when querying Pinecone vs. Quadrant (in terms of chatbot response times)? I'm currently all set up on Pinecone so migrating isn't ideal but if Quadrant is way faster, I'd for sure make the move.
Also, I'm hoping for Voiceflow to stream it's chatbot responses (to improve the UX), but I'm guessing that's not possible when we make calls to Flowise?
Appreciate it again! New subscriber over here
I would say they are probably very close, probably what matters more is the rest of your stack, here a difference between them probably 0.1 seconds so not a big deal. Well my workaround with streaming is just try to limit the length of responses this way the answer come faster, so instead of one step with long response truncate it in two smaller steps to avoid long lag. Also if you decide that you need streaming you can always just go full code and build it on NextJS
Awesome, thanks a lot for your perspective. Really helpful all around@@aisimp
@@grindwithmal no worries, feel free to join the discord I am making a collection of the best tech/tutorials in the ai space for developers, I am not biased 😂 I mention other UA-camrs/bloggers if they are good
Sounds good!@@aisimp
Thanks for the content, it's very valuable ! I have 2 questions : where do you host flowise ? Is it way cheaper to go with flowise ?
I'm hosting flowise on AWS free tier, its super fast, better than using Render. Bit more complicated to setup but there is a guide on Flowise Docs....
More tutorials! You're a god
On it! 🙏🏻🤗
🔥🔥🔥🔥 keep it coming
Will do right after I am done ✅ with the tutorial
Where is the course?
Building it 🤖
@aisimp Hi there! Great content dude. Some quick questions - Voiceflow has knowledge base so when does it make the case to use Flowise / Stack-ai? I want to make a training bot which references documents and URLs in a sharepoint - that is not scaleable on Voiceflow so what tech stack would you use and why?
If I wanted to reduce the cost of tokens whats a good LLM to use whilist maintaining accuracy?
Thanks in advance
EDIT: Just saw previous post had similar question but if anything anyone can add. VF alone can do a decent job but not scaleable I take it If I want to crawl 1000s of urls?
I think it is just easier to write a short Python widget that would loop through your list and chunk and upload all the content
Hello breat value ! Where is it possible to find a ressource that show how to paramrtrize this step by step please ?
Go to the discord, I have a link there under resources
My Flowise API call url is 'localhost:3000' and doesn't work when requested from VoiceFlow, how do i switch it to a domain like yours?
lol you are using it on a local machine, you need to expose your local host to external traffic. You gotta use ngrok
@@aisimp Yeah I'm watching a tutorial on how to put it on Render now, cheers
@@BRENDAN28 yeah running locally has its own technical perks you need to know
@@aisimp I have gotten it to run on Render, but after a few hours when I come back and reload the page, my project is no longer there. Have you experienced this?
@@BRENDAN28 they don’t have persistent storage. I added this module in my AI-EM30
Where can I learn more about your Course?
I am filming right now but bunch of ideas that I will have in the course I share in my AISIMP dev letter. In the future I can just reference each letter (they are numbered) to help with understanding.
Sign up link is in the description down 👇. Already put two posts no fluff, pure value
Is HITL possible with this stack?
Hello bro, thank you for this excellent content! Can you explain me how to easily get vectors in Qdrant? Thanks :)
the easiest way is to use flowise, create upsert model and using flowise platform feed it the files or you can use python and qdrant client
@@aisimp I think it's more easy with flowise 🤣 Thanks bro
I want to integrate my whatsapp cloud API CRM with Flowise AI solution with qdrant
Love this 🥰🥰. Curious how can we add this to Telegram to have TTS and STT features?
Funny that you ask, I connected Telegram bot to whisper to STT and used elevenlabs for STT. There is a slight delay because elevenlabs can’t export in OGG format so I format it from mp3 to OGG and then send it to Telegram, all that done in Flowise and Telegram is just the front instead of Voiceflow
Would love a telegram vid
@@musumo1908 sounds good, I think I will be accumulating all of these prototypes somewhere cause editing these videos takes forever and the video editor is a secret 🤫 … no one is sharing it, found it on twitter
Flowise or stack ai which platform is better
Whichever is opensource
Dude you already are pretty well worse in flowise Ai and stack Ai is also mostly same its really confusing which one to chose. Can you make a video on it
@@siddhantgupta1300 I find Flowise better in every way, for a start its more user friendly, its open source, and its of course a lot cheaper....
congratulation, very interesting content. What about Uchat?
Never used so have no idea
Nice and informative video!
Can someone please explain why @aisimp said that this setup is cheaper token-wise than setting up directly on Voiceflow? Around 18:11
Voiceflow is using Openai for AI functionality and then charges you 2.5x more cost wise vs just using Openai directly and Flowise is free. So here is the reason for that statement
Thanks@@aisimp so besides the fee for the platform, 2.5x more? Wow. Thank you very much.
Not really.m Voiceflow fee is calculated based on the usage of Ai Tokens, but if you run out of your plan then you need to upgrade or buy more at that extra price. People pay because voiceflow appears easier to use, other than that it is on parity with many other applications@@RolandoLopezNieto
I get it now, thank you very much, this is an awesome video, waiting for more and the course you mentioned, thanks again.
This is why I simp for you
😂😂😂 got me to chuckle 🤭
nice
how can i delete memory in flowise ?
Flowise has no memory, there are many memory modules that you can interchange based on the use case
I've had trouble with setting up flowise last time. It wouldn't save the model I built.
and can't find the discord link anywhere
Will post it soon also check out my previous flowise video and in the discord I have a long video on how to setup flowise in render or railway
@@aisimp thanks buddy, will do
hey, could you make a more simpler video with FLOWISE+VOICEFLOW?
In what way simpler?
@@aisimp I don't quite understand those API integrations, intents, entities, those variables and so on. So I request you to create a simple one with both of them being used.
And a question ...
If I use flowise and suppose I only pay for GPT3.5 API, should I need to subscribe for voice-flow or not?
And why use both? Can't we do this whole thing from FLOWISE only?
@@suyogkhanal5542 I don’t think so, flowise chat UI is good but not as advanced and logic also has to be done a bit differently
Hii great video man, please is make webhook to google sheet free?
Yes MAKE’s webhook connection to google sheet is free up until 1000 executions per month
What about langflow?
Langflow is Python alternative to Flowise. Will have Langflow examples next week