Comment your thought on the potential of these GPTs! 👀👇🏽 📚 My Free Skool Community: bit.ly/3uRIRB3 🤝 Work With Me: www.morningside.ai/ 📈 My AI Agency Accelerator: bit.ly/3wxLubP
Liam, do you know if it is possible to use Voiceflow in Sandbox mode for a production chatbot? The pricing chart suggests that you need Pro for code steps, but are there any other steps in your example code that are Pro-only?
This is absolutely a mine gold!! So I'm a noob on Python and a wanabee AI master, and I tried following your video but got stumbled when I first run it.The json file wasn't cerated and there were some messages in the console (see below) Hope you could shed some light on this error: raise self._make_status_error_from_response(err.response) from None openai.BadRequestError: Error code: 400 - {'error': {'message': "The requested model 'gpt-4-1106-preview' does not exist.", 'type': 'invalid_request_error', 'param': 'model', 'code': 'model_not_found'}} 😕
I'm the creator of the Startup Mentor agent that showed up on one of your slides. Thanks for the plug, but even more, thank you for showing me how to spread this app outside of GTP-4. BTW, I may be your oldest subscriber: I'm an 81 year old Vietnam Vet and former lawyer, and I've got a lot to learn from you.
Hold on there, partner! I'm a 81 year old consulting mechanical engineer creating applications for building energy management. (I was in the Army too but managed to skip Vietnam.) Like you I've got a lot to learn!
I would like to extend two big thank you's to you.... The first for your service, and the second for showing the world that there are rules or age restrictions on learning new things. 🙌
you are extremely generous giving away these steps and this code. I am a software engineer/ ai architect, and am thoroughly impressed you gave so much value and agree the money saved doing this stuff is HUGE compared to what this would cost on AWS if someone doesn't know how to use flask and postman and implement LLMs with vector databases for larger knowledge bases. For general websites this is great, and I still maintain there is acceptable security only when you host in something like aws for people who have more sensitive data. This is also the same thing as using wordpress ai engine, which is also very cost efficient once set up. The biggest value I always take away from your videos is how to turn this into a business, many thanks!
Great step-by-step guide, we use something very similar in our builds. What I would mention is that the context length (due to long-term chat history) could become very expensive if not managed/programmed correctly. For short term memory we typically limit this to the 5 past interactions, this helps reduce the token count. For longer term we use a VectorDB that saves any conversation. It's still coherent regardless of which method it uses and it keeps the token count way down which can become easily inflated.
short term / long term - what you mean by that? like different use cases for different clients? like: it depends on customer journey? if someone has hard decision to make, some good is not cheap, etc. vs something easy when it comes to decision? that's what you mean? thanks for this comment
@@simarpreetkaur8674 Yes, OPENAI API is pay to use. You'll want to consider financially if you want to use it and also protect yourself from attackers that'll spam use it.
Liam, I really appreciate u sharing this info. I'm in no way a coder, but love the tech Magic!! U do an excellent job of breaking it down for the rest of us. But please be careful when offering use cases, cuz that is the real point of understanding what a product can be and how it might work for a client. In the example in this video, u gave an excellent explanation of how the back end is built, apps u used to build, and the pre-designed coded ur offering for free, brilliant (and kind). At the end, it appeared u were building an internal chat for in-house salespeople but added it to a consumer-facing site. Let me know if I missed something. I'm just trying to wrap my head around it all. Much love and gratitude. Alisa
I just wanted to let you know Liam that you are an amazing person and I love you from one human to another. You are a great inspiration. Whenever I watch your content all I hear in my mind is, if you help enough people get what they want, you end up getting what you're after as well. Keep fucking going bro. You're appreciated.
You are the answer to my prayers, I am so grateful. Thank you and I am really excited to learn so much more from you. I'm starting a new career path at 47 years young re AI. Bless you!
Good video. Some feedback, you kept talking about your resource hub in the description, but if you ctrl-f on your description "resource" doesnt list once. Instead you have it listed in a pinned comment. Took me a minute to find it, which could lose some conversions.
not bad for non technical folks. Especially, if someone doesn't want to do much programming or struggles with it. Also, does a fair job of explaining the mechanics for piecing/hacking together a chatbot for semi-technical folks.
How do I get your resources? I try to Sign Up to Agentive and it tells me I have to be in a waiting list. It is frustrating not to be able to get instant access. Am loving the video and how you simplify the steps. Great Job
Thank you that you explained perfectly the differences between creating GPTs on ChatGPT and using the API and their own functionalities. It should have been better if you had explored more on the possibilities of ChatGPT actions also, but we can do it on another video. 🙏👍
@@chrisbeechey9841 interesting question. I have created GPTs, but I always have to instruct them that is they can't find the answer either in the knowledge base or other means, then search the internet, but I think this instruction can be placed definitely on the bot settings.
@@chrisbeechey9841 I think you can add some Research Agent that will scrape websites, maybe with langchain, might be used with 3.5 turbo I saw something like that on AI Jason's channel
Bro you are such a legend for sharing this and giving resources too . Thank you so much this is very exciting. Just made an account and cant wait to make something. Bless you my man. Keep up the good work.
Thanks for the tutorial, I created for own E-learning business and collected leads to google sheet. From google sheet I connected to whatsapp. Thanks again.
Why would you use replit to create an assistant with a knowledge base rather than only voiceflow and use the voiceflow knoweldgebase feature? What's the benefit to the added tool/step? Thanks so much for the awesome video!
Same here. Was scanning all the comments for this very question. From what I have seen in VoiceFlow you can do just about anything including advanced RAG and fine-tuning. Why write all the code in Replit? In another video you said you use StackAI in addition to VoiceFlow when more complex code is needed. It's hard to figure out which videos to learn from (in general with all AI TH-cam channels) as the space is moving so quickly. I almost don't want to watch videos are over a month or two old. Would love some feedback on what are the best practices / tools to build advanced chatbots. Thanks!
@@mikerichert The biggest benefit is that you can use your own API keys and are therefore no longer subject to Voiceflow's token limits. The AI calls are therefore billed externally. I don't know if it's cheaper in the end. It's just stupid that Voiceflow has a 2M token limit with the Pro version for 50 dollars and the next level is enormously more expensive. However, I have no idea how much you have to run to reach 2M tokens.
Hey Liam thank you again for all the amazing value you provide. I have just one question. What is the difference between using the assistant API vs just uploading our custom knowledge base into voiceflow? The end product on the website looks the same after all
1. OpenAI knowledge base is way better 2. This tut isn’t including custom functions which when setup correctly in the assistant unlocks massive potential This should more be seen as a migration over towards using OpenAI as the entire back end
@@LiamOttleythank you for sharing this info first, across the world. I will join your community now. You rock! - the team here sends you greetings from Edinburgh!
One question: how can I prevent this from being overloaded by trolls or bots? Like if I put it on the website, I pay per usage of the API. So if a bot or a person just sends a lot of requests - can I limit the amount of requests per IP per hour?
Looks like its going to be botpress for me. I am 5020 on the agentive waiting list. Informative video however and I might me your second oldest viewer lol. Grabbed a sinclair and commodore 64 at age 16 was hooked ever since. Being a systems consultant for decades I am loving the AI rabbithole I have found myself in.
God bless you. I am shocked by how much value you give for free. It feels so illegal to know your channel. I hope you achieve your goals with this channel !
Voice flow is a pretty cool way to quickly have a front end design for a chatbot thanks for revealing that. I’m guessing you could still replicate the backend on Vs code as I’m not too sure what set up / editor you was using there. Separately, could you cover this for big organisations, like connecting the assistants api to a database rather than a few local files because with huge companies they’re more likely to have a ton of info on a database and I think with assistance api there’s like a 20 file limit upload? Correct me if I’m wrong.. nevertheless just subscribed. Another great channel.
Exactly what I was hoping was possible! Not sure if I can grasp all this myself to being into my website however. I hope some easier ways are not far off
This is game changing I wanted to build this internally at my company so glad we don’t have to now ! Question is there a limit to the size or number of knowledge items you can upload to the retrieval aspect of the assistants api ?
@@dads-weekend Ah interesting I thought that since the knowledge is handled via file upload it would be different since it's a retrieval (RAG) mechanism. I am asking more about the storage of n number of files. The 120k token window applies to when the knowledge is retrieved and injected into the prompt, but I'm not sure that applies to dormant (not injected) knowledge.
I just got to know about voiceflow through this video. Does it serve any purpose other than providing the UI component? In my understanding, everything is being done by the assistant. Right?
❤ 00:01 Learn how to add custom GPTs to any website and understand the value of GPTs in the real world. 01:46 Create a custom knowledge chatbot using the Assistant API. 03:21 Junior sales reps can use a custom chatbot to get information about the business. 04:51 The main.py file contains API endpoints for communication. 06:22 To add custom GPTs to any website, follow these steps: 07:49 Custom GPTs can be easily added to any website using Repeat template 09:17 Setting up custom GPTs on a website 11:04 Adding custom GPTs to any website in minutes using Voiceflow 12:42 Creating a custom GPT using voice flow and the assistant API for websites Crafted by Merlin AI.
How can I get the code for the clone info? It's saying I need an Agentive account and it's been several days I have been waiting and I am on a deadline.
Great video as usual. 😀 I have a question, after a customer has decided use your chatbot service, do you charge them a retainer for future access to your business an if so at what price point? Yearly or monthly charge? What do you do?
Liam, thank you so much for the valuable info. You broke things down really well for those of us who are not so technical (like me.) Question: if I download Assistants API and follow the steps indicated, will I then have to train the bot? Or does it come out of the box trained on GTP4 data sets?
But Liam, won't replit timeout the chatbot after being idle? Also you would need your computer on at all time to keep the replit code active and running. Can you do a tutorial that makes it run a traditional server where it has permanent functionality instead of demo? Thanks mate
One question @LiamOttley. Is it possible to create and add a custom knowledge base in chat gpt builder and than access it via Api? Or do we have to config it in replit?
Yes, I have this question too. Would love to train a custom GPT to use internally by our team with our companies specific technical knowledge, and eventually graduate this to a chat bot on our website
Any update on this? Correct me if Im wrong but with the knowledge.docx does this need to be created before you run the Assistant on Replit? After it has been created is there a way to update or add more knowledge to the chatbot?
The main.py file you are showing in the video doesn't look anything like the one you have linked in the description. I'm lost because I'm not seeing the same thing.
Thanks! This video is just what I needed! Quick question, how do I give weblink browsing capabilities to my assistant? Like how I can check the box for browsing capabilities while creating a custom GPT on Chat-GPT, is there a way to enable it in the code provided in riplet? Thanks
Awesome stuff mate! Thanks for the information. Is it possible to do something like this without the OpenAI key and Riplet? For example if a customer would like to run it locally.
can you do a long term form with updates with all the new stuff coming up recently ? how to build an chatbot with all the newest version ? would be great Liam ! 🤖
This all looks great. First custom application of the Assistants API I've seen. In order to get it working on a real website, would I also need to purchase a paying plan on Replit and Voiceflow?
Very nice video! Thank you. Im new to coding and doing my bachelors. As a project i want to make a web app that has a AI chatbot, but does it cost me anything if i do it this way?
So I got this working on my website, thanks a ton! Im a complete noob when it comes to web design and setup and coding. I have a question though, Is there a way to have a record of each interaction with the chatbot sent to my email or somewhere else? I want to see the chat bot side and the user side if possible.
@@Neotech1234 Im still working on it but there is a chat log inside the voice flow web app. If you look on the left hand side of the dash board there's a conversation bubble icon and in there is where the conversations are stored. You can see them there. Which is good, since now I just have to figure out how to get those conversations sent to my email.
Hi Liam, that is great! Thanks a lot. Only unfortunately I cannot get it working...:(. Two questions: 1. I did not see you using the 'Deploy to production' function in Replit, is that not necessary? 2. I watched the video many times (and followed step by step), but I cannot find how/where to connect the assistant ID to Voiceflow at all? Please help, I'd love to get this working please, thanks a lot
What changes will be needed (if this is at all possible) if I have already created a GPT in open AI and all I want to do is to provide an interface to that GPT from my website (so they do not a chatGPTPlus account) ? From what I know currently it can only be accessed if my users have a chatGPTPlus account and I provide them the GPT url.
This isn't currently possible so you'd need to create an Assistant and use the API to surface the GPT. You could copy/paste instructions from the custom GPT into your assistant
Love the video! GPTs are awesome and use cases are endless. Already created a few for fun. Think our chatbot builder platform has more functionality. But we might be bias 🤣
Liam, thank you a lot for this great content! It is much appreciated! I succeeded in setting this up for myself. But no the speed for the answers do need quite a long time. Anyone has a tip on howo to speed the chatbots ansers up?
Another great video Liam. Question though. You don’t think it’s worth creating a good app for the OpenAI App Store that would do well long term when thousands to maybe be millions use it and we get paid off of token use? I am currently learning both but to my embarrassment I learned about the api sight of OpenAI about 3-4 days ago and have been playing around in it.
Liam, in recent video you did you mentioned the importance of being able to market and sell but also to BUILD. It seems you break down all in great detail, but you also seem to reference a fair amount of integrative code. To enter this business is there code I should be learning as well. Fortunately I did a lot o FORTRAN when I was young in 1979, so I understand the logic flow etc.
@@boukm3n how long do you think it would take me to learn an adequate amount of python to really be able to do what I need to do in building ChatGPT’s? Assuming I really put some time and effort into it
Excellent video! Just one question: Whats the cost structure to run a project like this? I have to pay for VoiceFlow and also for the ChatGPT API? Or just Voiceflow? Could you clarify this please?
This is a start in understanding how to implement a custom no code/low code chatbot but after reading some of the comments it seems there could be some unpleasant surprises revealing themselves when the usage bill arrives.
Can you have a chat assistant bot for something like a coaching program that clients can ask it questions while you are out of office? Also, how do you load up your knowledge base?
Comment your thought on the potential of these GPTs! 👀👇🏽
📚 My Free Skool Community: bit.ly/3uRIRB3
🤝 Work With Me: www.morningside.ai/
📈 My AI Agency Accelerator: bit.ly/3wxLubP
Liam, do you know if it is possible to use Voiceflow in Sandbox mode for a production chatbot? The pricing chart suggests that you need Pro for code steps, but are there any other steps in your example code that are Pro-only?
Just a FYI @Liam.... this guy is ripping off your content: th-cam.com/video/rbc4qDuxhes/w-d-xo.html
resource link not working for me
This is absolutely a mine gold!! So I'm a noob on Python and a wanabee AI master, and I tried following your video but got stumbled when I first run it.The json file wasn't cerated and there were some messages in the console (see below) Hope you could shed some light on this error:
raise self._make_status_error_from_response(err.response) from None
openai.BadRequestError: Error code: 400 - {'error': {'message': "The requested model 'gpt-4-1106-preview' does not exist.", 'type': 'invalid_request_error', 'param': 'model', 'code': 'model_not_found'}}
😕
The clone option is missed
I'm the creator of the Startup Mentor agent that showed up on one of your slides. Thanks for the plug, but even more, thank you for showing me how to spread this app outside of GTP-4. BTW, I may be your oldest subscriber: I'm an 81 year old Vietnam Vet and former lawyer, and I've got a lot to learn from you.
Hold on there, partner! I'm a 81 year old consulting mechanical engineer creating applications for building energy management. (I was in the Army too but managed to skip Vietnam.) Like you I've got a lot to learn!
@@michaellavelle7354 LOL we all do I am approaching you.
amazing look at you guys go :)
Wow!!
I would like to extend two big thank you's to you.... The first for your service, and the second for showing the world that there are rules or age restrictions on learning new things. 🙌
you are extremely generous giving away these steps and this code. I am a software engineer/ ai architect, and am thoroughly impressed you gave so much value and agree the money saved doing this stuff is HUGE compared to what this would cost on AWS if someone doesn't know how to use flask and postman and implement LLMs with vector databases for larger knowledge bases. For general websites this is great, and I still maintain there is acceptable security only when you host in something like aws for people who have more sensitive data. This is also the same thing as using wordpress ai engine, which is also very cost efficient once set up. The biggest value I always take away from your videos is how to turn this into a business, many thanks!
🧢🧢
@@devmehta7406why do you think they’re lying ?
Great step-by-step guide, we use something very similar in our builds. What I would mention is that the context length (due to long-term chat history) could become very expensive if not managed/programmed correctly. For short term memory we typically limit this to the 5 past interactions, this helps reduce the token count. For longer term we use a VectorDB that saves any conversation. It's still coherent regardless of which method it uses and it keeps the token count way down which can become easily inflated.
short term / long term - what you mean by that? like different use cases for different clients? like: it depends on customer journey? if someone has hard decision to make, some good is not cheap, etc. vs something easy when it comes to decision? that's what you mean?
thanks for this comment
Can u pls clear my 1 doubt, is OPENAI_API_key paid?
Please reply asap, I am really confused, I have tried alott but I always get billing errors in vscode
@@simarpreetkaur8674 it is. api key from openai is paid
@@simarpreetkaur8674 Yes, OPENAI API is pay to use. You'll want to consider financially if you want to use it and also protect yourself from attackers that'll spam use it.
Liam, I really appreciate u sharing this info. I'm in no way a coder, but love the tech Magic!! U do an excellent job of breaking it down for the rest of us. But please be careful when offering use cases, cuz that is the real point of understanding what a product can be and how it might work for a client. In the example in this video, u gave an excellent explanation of how the back end is built, apps u used to build, and the pre-designed coded ur offering for free, brilliant (and kind). At the end, it appeared u were building an internal chat for in-house salespeople but added it to a consumer-facing site. Let me know if I missed something. I'm just trying to wrap my head around it all. Much love and gratitude. Alisa
Thanks Liam!! I owe you a lot, finally I managed to get my GPT Assistant available on my webpage for any user to use.
Can you provide the website please
You need to do a video on pricing based on usage so we can properly lay out a budget for our clients
swap your first 2 words with Could you, maybe say thank you...he might just do that :)
I just wanted to let you know Liam that you are an amazing person and I love you from one human to another. You are a great inspiration. Whenever I watch your content all I hear in my mind is, if you help enough people get what they want, you end up getting what you're after as well. Keep fucking going bro. You're appreciated.
You are the answer to my prayers, I am so grateful. Thank you and I am really excited to learn so much more from you. I'm starting a new career path at 47 years young re AI. Bless you!
Have you decided to learn to code?
@@Jamesrwatsonx yes! 😁👋
Thanks! I appreciate being cited as the top comment asking for this!
Good video. Some feedback, you kept talking about your resource hub in the description, but if you ctrl-f on your description "resource" doesnt list once. Instead you have it listed in a pinned comment. Took me a minute to find it, which could lose some conversions.
Great fast review on the project. Looking forward to that full version Liam 👍
not bad for non technical folks. Especially, if someone doesn't want to do much programming or struggles with it. Also, does a fair job of explaining the mechanics for piecing/hacking together a chatbot for semi-technical folks.
How do I get your resources? I try to Sign Up to Agentive and it tells me I have to be in a waiting list. It is frustrating not to be able to get instant access. Am loving the video and how you simplify the steps. Great Job
same here. looking a way to access the resources but being told that there is a thousands of ppl ahead of me in the waiting list. Please help
These are actually some sick videos, thanks for sharing!
This is really cool. as I was watching I saw the subs hit 107k. Well Done! as usual
Thank you Liam for putting out this video . I commented on your last video asking about the API , this answers my question . Tremendous Value❣
Thank you that you explained perfectly the differences between creating GPTs on ChatGPT and using the API and their own functionalities. It should have been better if you had explored more on the possibilities of ChatGPT actions also, but we can do it on another video. 🙏👍
Left em out on purpose to not over complicate this concept for beginners. Another video coming showing custom actions!
With OpenAI Assistant, can you create chatbots that pull answers from the web? Not just uploaded files?
@@chrisbeechey9841 interesting question. I have created GPTs, but I always have to instruct them that is they can't find the answer either in the knowledge base or other means, then search the internet, but I think this instruction can be placed definitely on the bot settings.
@@chrisbeechey9841 I think you can add some Research Agent that will scrape websites, maybe with langchain, might be used with 3.5 turbo
I saw something like that on AI Jason's channel
You're my Guru dude! thanks for sharing
Brilliant. Thank you for the education.
Thanks Liam! Great content
Hey bro, I went to Dilworth and I'm in the Staples Program. Enjoyed watching your story!
Great course dude! That's a lot of value for free!
0:51 Ayyyye that’s me! Great video again Liam
Thank you! Your videos are amazing, i work in a different area, but i love ai, and this give me a great idea to use it in my field!!
very helpful tutorial, thanks for sharing!
Bro you are such a legend for sharing this and giving resources too . Thank you so much this is very exciting. Just made an account and cant wait to make something. Bless you my man. Keep up the good work.
Thanks for the tutorial, I created for own E-learning business and collected leads to google sheet. From google sheet I connected to whatsapp. Thanks again.
Pure value! 🔥
Thank you, bro'! 🙏🏼
Why would you use replit to create an assistant with a knowledge base rather than only voiceflow and use the voiceflow knoweldgebase feature? What's the benefit to the added tool/step? Thanks so much for the awesome video!
I also want to know the same thing, are there specific benefits of using replit in combination with voiceflow?
Same here. Was scanning all the comments for this very question. From what I have seen in VoiceFlow you can do just about anything including advanced RAG and fine-tuning. Why write all the code in Replit? In another video you said you use StackAI in addition to VoiceFlow when more complex code is needed. It's hard to figure out which videos to learn from (in general with all AI TH-cam channels) as the space is moving so quickly. I almost don't want to watch videos are over a month or two old. Would love some feedback on what are the best practices / tools to build advanced chatbots. Thanks!
@@mikerichert The biggest benefit is that you can use your own API keys and are therefore no longer subject to Voiceflow's token limits. The AI calls are therefore billed externally. I don't know if it's cheaper in the end. It's just stupid that Voiceflow has a 2M token limit with the Pro version for 50 dollars and the next level is enormously more expensive. However, I have no idea how much you have to run to reach 2M tokens.
Super video Liam, is there a way to make the script scrape a website it is to be used on instead of uploading a document?
Very nice! Thanks man! Im gonna set it up for my real estate business!
Thank you, master, for your sharing. I have learned a lot
Thanks Liam for this vidz that was so ething I was trying to figure out
Hey Liam thank you again for all the amazing value you provide. I have just one question. What is the difference between using the assistant API vs just uploading our custom knowledge base into voiceflow? The end product on the website looks the same after all
1. OpenAI knowledge base is way better
2. This tut isn’t including custom functions which when setup correctly in the assistant unlocks massive potential
This should more be seen as a migration over towards using OpenAI as the entire back end
@@LiamOttley okay that makes sense thank you! Your next video will go more in depth on how to add custom functions right?
@@LiamOttleythank you for sharing this info first, across the world. I will join your community now.
You rock! - the team here sends you greetings from Edinburgh!
can a Custom GPT have excess on my own website?🙊
@@fabio-y5qnope
One question: how can I prevent this from being overloaded by trolls or bots? Like if I put it on the website, I pay per usage of the API. So if a bot or a person just sends a lot of requests - can I limit the amount of requests per IP per hour?
yes I would like to know this too!
Same
you can try cloudflare rate limiting
My guy! You are literally the man for this. And your timing is spot on 🎯 🔥
Hi Liam, super excited for this AI chatbot solution, but does this mean we expose our internal knowledge document to openAI or the public?
13:00 lovely! can't wait
Looks like its going to be botpress for me. I am 5020 on the agentive waiting list. Informative video however and I might me your second oldest viewer lol. Grabbed a sinclair and commodore 64 at age 16 was hooked ever since. Being a systems consultant for decades I am loving the AI rabbithole I have found myself in.
God bless you. I am shocked by how much value you give for free. It feels so illegal to know your channel. I hope you achieve your goals with this channel !
Voice flow is a pretty cool way to quickly have a front end design for a chatbot thanks for revealing that. I’m guessing you could still replicate the backend on Vs code as I’m not too sure what set up / editor you was using there. Separately, could you cover this for big organisations, like connecting the assistants api to a database rather than a few local files because with huge companies they’re more likely to have a ton of info on a database and I think with assistance api there’s like a 20 file limit upload? Correct me if I’m wrong.. nevertheless just subscribed. Another great channel.
Mate, you earned my attention. Thanks and eill get back here.
Can’t thank you enough for this break down!
Huge bro - You've just opened the flood gates.
Building like crazy right now 💪
Hey, just looking for the template to download. That's where I'm stuck. Great videos - thanks so much!
I have no idea how to access the resource hub or clone the replit. Can anyone help? 🤔
(First time I have heard the term 'replit', actually...)
Great stuff. Thanks
Exactly what I was hoping was possible! Not sure if I can grasp all this myself to being into my website however.
I hope some easier ways are not far off
This man is dropping gold
Liam - well done!
This is game changing I wanted to build this internally at my company so glad we don’t have to now ! Question is there a limit to the size or number of knowledge items you can upload to the retrieval aspect of the assistants api ?
I would think it's the new token limit of 120,000 tokens. Can anyone confirm that?
@@dads-weekend Ah interesting I thought that since the knowledge is handled via file upload it would be different since it's a retrieval (RAG) mechanism. I am asking more about the storage of n number of files. The 120k token window applies to when the knowledge is retrieved and injected into the prompt, but I'm not sure that applies to dormant (not injected) knowledge.
I'd like to know too
Same question
Great job. Liam, how big is your production team? Results are great!
Thanks for the info
you read me.. literally thinking about it 😂 thank you
You earned a subscriber
Hey Liam Amazing Video, where did you learn how to code?
bro coming with the coldest content
I just got to know about voiceflow through this video. Does it serve any purpose other than providing the UI component?
In my understanding, everything is being done by the assistant. Right?
That is pretty cool. Thank you Liam. Would this bot have a long term memory? I mean are threads saved across sessions?
Amazing, thanks!
❤
00:01 Learn how to add custom GPTs to any website and understand the value of GPTs in the real world.
01:46 Create a custom knowledge chatbot using the Assistant API.
03:21 Junior sales reps can use a custom chatbot to get information about the business.
04:51 The main.py file contains API endpoints for communication.
06:22 To add custom GPTs to any website, follow these steps:
07:49 Custom GPTs can be easily added to any website using Repeat template
09:17 Setting up custom GPTs on a website
11:04 Adding custom GPTs to any website in minutes using Voiceflow
12:42 Creating a custom GPT using voice flow and the assistant API for websites
Crafted by Merlin AI.
How can I get the code for the clone info? It's saying I need an Agentive account and it's been several days I have been waiting and I am on a deadline.
Great content! For production, is Replit Core plan enough or do I use Reserved VM?
very good , thanks
Great video as usual. 😀 I have a question, after a customer has decided use your chatbot service, do you charge them a retainer for future access to your business an if so at what price point? Yearly or monthly charge? What do you do?
Great question
I'd love to know this too
Hi Liam,.This is incredible.You great man! Is it only for IOS? Can it works on Android?
nice video, thanks
Can you do this in a WordPress website?
Liam, thank you so much for the valuable info. You broke things down really well for those of us who are not so technical (like me.) Question: if I download Assistants API and follow the steps indicated, will I then have to train the bot? Or does it come out of the box trained on GTP4 data sets?
But Liam, won't replit timeout the chatbot after being idle? Also you would need your computer on at all time to keep the replit code active and running. Can you do a tutorial that makes it run a traditional server where it has permanent functionality instead of demo?
Thanks mate
Perfecto para probarlo hoy. Gracias te cuento luego que tal me fue. 👌🏽
1:50 this is exactly why I found this video. Got unto custom gpts but didnt know the next steps. Now I do, the assistant api
One question @LiamOttley. Is it possible to create and add a custom knowledge base in chat gpt builder and than access it via Api? Or do we have to config it in replit?
Yes, I have this question too. Would love to train a custom GPT to use internally by our team with our companies specific technical knowledge, and eventually graduate this to a chat bot on our website
Any update on this? Correct me if Im wrong but with the knowledge.docx does this need to be created before you run the Assistant on Replit? After it has been created is there a way to update or add more knowledge to the chatbot?
The main.py file you are showing in the video doesn't look anything like the one you have linked in the description. I'm lost because I'm not seeing the same thing.
Make sure you’re clicking on the right section of the resource hub, the name needs to match this video
hey liam i cant find the resources you told me are in description
Thanks! This video is just what I needed! Quick question, how do I give weblink browsing capabilities to my assistant? Like how I can check the box for browsing capabilities while creating a custom GPT on Chat-GPT, is there a way to enable it in the code provided in riplet? Thanks
is replit and voiceflow services to create this project free?
Awesome stuff mate! Thanks for the information. Is it possible to do something like this without the OpenAI key and Riplet? For example if a customer would like to run it locally.
I like your hairstyle man
can you do a long term form with updates with all the new stuff coming up recently ? how to build an chatbot with all the newest version ? would be great Liam ! 🤖
This all looks great. First custom application of the Assistants API I've seen. In order to get it working on a real website, would I also need to purchase a paying plan on Replit and Voiceflow?
Very nice video! Thank you.
Im new to coding and doing my bachelors. As a project i want to make a web app that has a AI chatbot, but does it cost me anything if i do it this way?
Awesome Liam!!
🔥 Information! Quick question, can this be used on a mobile web app?
So I got this working on my website, thanks a ton! Im a complete noob when it comes to web design and setup and coding. I have a question though, Is there a way to have a record of each interaction with the chatbot sent to my email or somewhere else? I want to see the chat bot side and the user side if possible.
Me too! I tried but can't get it to work...
@@Neotech1234 I’m going to start working on that functionality now, hopefully I can make it work.
@@Neotech1234 Im still working on it but there is a chat log inside the voice flow web app. If you look on the left hand side of the dash board there's a conversation bubble icon and in there is where the conversations are stored. You can see them there. Which is good, since now I just have to figure out how to get those conversations sent to my email.
Hi Liam, that is great! Thanks a lot. Only unfortunately I cannot get it working...:(. Two questions:
1. I did not see you using the 'Deploy to production' function in Replit, is that not necessary?
2. I watched the video many times (and followed step by step), but I cannot find how/where to connect the assistant ID to Voiceflow at all?
Please help, I'd love to get this working please, thanks a lot
I also was unable to get the code from Replit, if you could help on this @liamottley it would be much appreciated
you use the URL of the server
@@alfiebutler20 thanks. i can see the server URL/assistant ID, I just have no idea where I need to plug this in/copy it to?
maybe somebody who successfullyi managed to get it working can help please? I have no tech skills and am lost... tks a lot
I also cant retrieve the url from replit, which doesnt open webview when running in replit, so cant find ID, can someone help?
Unfortunately I did not receive an e-mail with the code.😰
What changes will be needed (if this is at all possible) if I have already created a GPT in open AI and all I want to do is to provide an interface to that GPT from my website (so they do not a chatGPTPlus account) ? From what I know currently it can only be accessed if my users have a chatGPTPlus account and I provide them the GPT url.
This isn't currently possible so you'd need to create an Assistant and use the API to surface the GPT. You could copy/paste instructions from the custom GPT into your assistant
Love the video! GPTs are awesome and use cases are endless. Already created a few for fun.
Think our chatbot builder platform has more functionality. But we might be bias 🤣
Liam, thank you a lot for this great content! It is much appreciated! I succeeded in setting this up for myself. But no the speed for the answers do need quite a long time. Anyone has a tip on howo to speed the chatbots ansers up?
This is great! One things it still doesn’t do though is generate photos on demand like gpt. I have tattoo clients who would k!ll for that option
Hey Liam! Very useful thank you. Is it possible for API assistants to browse the internet? Many thanks :)
Another great video Liam. Question though. You don’t think it’s worth creating a good app for the OpenAI App Store that would do well long term when thousands to maybe be millions use it and we get paid off of token use? I am currently learning both but to my embarrassment I learned about the api sight of OpenAI about 3-4 days ago and have been playing around in it.
Liam, in recent video you did you mentioned the importance of being able to market and sell but also to BUILD. It seems you break down all in great detail, but you also seem to reference a fair amount of integrative code. To enter this business is there code I should be learning as well. Fortunately I did a lot o FORTRAN when I was young in 1979, so I understand the logic flow etc.
I think we came to the right place
Python
You should also know javascript, if you need help I'll happily help you
@@boukm3n perfect thank you. I will begin looking into it.
@@boukm3n how long do you think it would take me to learn an adequate amount of python to really be able to do what I need to do in building ChatGPT’s? Assuming I really put some time and effort into it
Excellent video! Just one question:
Whats the cost structure to run a project like this?
I have to pay for VoiceFlow and also for the ChatGPT API? Or just Voiceflow?
Could you clarify this please?
Is retune a more cost effective all in one solution?
This is a start in understanding how to implement a custom no code/low code chatbot but after reading some of the comments it seems there could be some unpleasant surprises revealing themselves when the usage bill arrives.
Can you have a chat assistant bot for something like a coaching program that clients can ask it questions while you are out of office?
Also, how do you load up your knowledge base?
Thanks!!
Nice, did you by pass the 20 files max upload for training?