Create an AI Agent in 5 mins with No Code

hello hello welcome to the next workshop on part of the beloud and build AI Series today we have awesome from Nexus GPT part of us part part joining us for the the workshop today um an interesting topic how to create an AI agent in five minutes using no code and the workshop is going to be 60 minutes maybe um but the agent creation part will be five minutes exactly yeah thank you very much uh Poria uh and yeah uh thank you everyone uh for joining this uh this session uh so as pora said my name is SM I built a cool tool called Nexus GPT uh and Nexus essentially is a platform that enables you to create AI agent without coding um we're going to go into the specifics of uh what I mean by AI agent and what Nexus is used for and we're going to go into a practical case as well but uh before going into that what I want to share with you is really kind of a bit more context on the AI market and why building agent like makes sense right now A lot of people are talking about it uh and most people don't really know what an AI agent is and what's really the difference between AI agent and and AI assistant so I'll try to kind of make it uh as simple as possible um going into the context and all of that and then we'll go into a practical case and of course por feel free to interrupt me if you have like any question or if like people in the audience have specific questions that I can answer during the presentation absolutely absolutely okay there there was an echo absolutely um Austin just before that bohar has prepared something we'll go over for two minutes if that's okay with you we'll pass it on to you um read the private chat messages I'm sending you as well so uh top right don't forget to press that yes exactly here thank you thank you for joining us uh Poria forgot to uh let me introduce myself and he just kept talking um I'm Bahar co-founder of beloud we are so happy that we have you here thank you for the time that you spending awesome uh and thank you for joining us so I'm gonna do a little bit of very short presentation about what are we doing why we are hosting this thing that we are hosting right now and what it's all about uh first of all hello hello everyone for joining us from on LinkedIn YouTube Twitter hello from everywhere thank you awesome for being here for the time that you were spending thank you audience and the question is what what do we do and who we are um me and pora we're co-founders of beloud beloud is an off for-profit organization helping startups growing raising um raising fund growing their customers to 10K uh we have an accelerator community in studio we have community of Founders Builders creators and investors in the accelator is where we help startup to grow the studio is where we build startups and we host programs and workshops such as this for our community members we are on a mission to import the next generation of Builders investors and creators into the imagination age but the question is what is Imagination age well if you ever heard of chatting sending emails um Facebook and and in those in those times that that was an Information Age and information age is in the past we are now in the imagination age where we have abundance of resources because of Technology such as AI arvr web 3 open AI chat gbt and these resources are enabling us to realize any ideas we have in mind and we don't have an excuse to not to realize the ideas we have we have everything we need so we were thinking now that you can go from prompt to anything in matters of seconds or minutes uh let's see if we can build a billion dollar startup in three days or not so we invited almost more than 300 people uh to challenge to challenge themselves and try to build a billion dollar startup in three days these ideas can be either in in AI or they can just leverage AI to build these startup ideas and in these projects so we created bad on build AI we started working on it for less than two two months ago and here we are it's going to be it's going to start in May 10 in um in a day or so you still have a little bit of chance to join in if you want to this is the QR code for that um thank you so much for being here follow us on social media to learn more about the upcoming workshops and the hackthon itself and the stage is yours awesome that's great thank you so much Bahar thank you um and I'll be sharing back my screen and that's it yeah it's sharing back perfect so yeah as I said um the the goal here is mostly to give you a quick demo on how you can do it uh but I want to go before that into the why you should be doing that and you know going to a bit more context and the specifics of what an AI agent is um before that a quick words um so the AI agent is like if you look at the market and you look at all these different Industries more and more of them are going to be disrupted by AI in the next couple of years so bahara introduced that uh in her talk on on beloud but I think it's really relevant and especially nowadays when it's easier than ever to create those uh those AI assistance those AI task and now those AI agents um so it's a big Market it's growing a lot it this are these are these figures that I'm showing you here are probably very conservative it's going to be much more than that like the the actual Market of servicing people through Ai and through potentially AI agents is going to be tripling in the next uh in the next decade um so what I mean by by it's a there is a big opportunity there's like three main drivers of this opportunity the first one is actually as I said the job exposure so if you look at all the job and all the tasks that a lot of people are doing um a lot of the tasks that people are doing come down to things that people write and those are the things that the llm are good at and what I mean by people things that people write it can be anything from you know writing analyzing documents or even translating or transforming documents right and if you think about it a lot of the things that we do in our day-to-day job and a lot of the things that people do in their day-to-day job has something to do with writing the output of that is writing and therefore you know the modern llm are really relevant to that so that's kind of the first driver the second thing is if you look at how much it costed to train a model those costs have been decreasing exponentially over the past few years so if you look at gpt3 the cost of building a gpt3 type model uh was a few million dollars uh um in in the past like in 2020 right when gpt3 came out and now when when you look at it a lot of people can build on top of those and build their own model top on top of Open Source model with just a few hundred dollars and they can reach or or go above Beyond a gpt3 level um large language model with just a few hundred dollars and a lot of data a lot of internal data uh and therefore like the third driver that is you know basically you have a lot of opportunity it costs a lot of very little money to create uh those things and the barrier to entrance the technical barrier to entrance is very low therefore you have a lot of solution that are emerging right and those solution they they basically in the way I'm seeing it they they they go into two different kind of categories the first one is the actual product and the second one would be the actual service right and let's dive into the service first because it's kind of the easiest to to explain um but in the service right uh before you had a lot of um consultant or agency that were focused on automation or on lean management and and basically uh the what what those agencies are doing right were doing was going into companies and finding inefficiency and helping companies becoming more efficient well the thing is right now with AI and with what I told you it's very easy for people to build this Automation and to build these a a these different AIS to go into companies and simplify task and automate task right so what you're seeing a lot right now in the market is you have a lot of what people call those triaa or AI automation agency which are relatively young people who have very good knowledge of what an llm can and cannot do and basically go into this company and Implement not only automation but agent uh to simplify the task of the workers right and the great thing with that is that it's on their side if you compare to all the other service industry so if you think about like not only Consulting but also marketing and all of that it's very simple for those people to show the value they're adding what I mean by that is if you go into a real estate uh a real estate company and you say hey I can uh I'm a marketing agency and I can make your your business grow with my Services it's very difficult to assess whether or not you're doing a good job it takes a bit of time and you can waste a bit of money but on the other and if if someone or an agency comes in and say hey I can simplify the task that you're doing and I can reduce your time doing these T by 20 30 50% it's very easy for them to show you and to show their clients what they're doing and how efficient it is so the sales part of the agency which is the hardest part right with any agency whether it's consulting or marketing or anything is simpler because it's easier for them to demonstrate the value they're adding and therefore this Market which is the AI agency market is growing a lot with it um the second type the second category in which is probably the category you're most interested in is really kind of the AI agent solution right so those are basically a agentic type of solution that are able to not only advise but also do task on your behalf right right and so you can think a lot of in know they are very wide in a lot of different categories right from sales from marketing from assistant all those are typically use cases where you could replace a low-level employee with an agent right and I can dive into each of those specific solution right so if you go into for example therapy C which is the simplest one it has been uh made by um the same guy who made nomat Le who is called Peter levels and this guy basically what he did is that he put a GPT over a prompt and he enabled that gbt to receive different voices and access different knowledge and he put it online and he basically enabled what we called the therapy AI right and so that therapy AI was able to respond to questions remember conversation and give advice people which is kind of like like coaching type of agent and and this thing is generating quite a bit of money for him uh he I I think he launched it something like uh a few weeks ago now it's doing something around like a few thousand dollar uh Mr which is quite substantial for something that is just a few weeks old um if you go below right uh this one here it's basically a solution that do screen calling right so you basically give an agent and you give it a number and and you transfer every call that you receive to that agent so the agent can uh can screen people who are calling you and decide whether or not they are uh solicitors scammers or actual people who want to talk to you and you know them um you also have like more complex type of agent which are you know either assistant type a agent who has access to uh tools like uh calendar emails uh and internet sech who are able to autonomously decide whether or not they should like book a meeting for you go and send an email to remind people of your upcoming meeting or anything basically what an assistant would do uh and you have like sales agent for example so yala is a company a startup that created a sales agent on WhatsApp so it's basically an agent that is trained on the data of a website or an e-commerce or any type of business and can not only answer questions but also prompt people to upgrade them and prompt people to buy more stuff from you so all of those are kind of use cases of Agents right but the if if you look at it on a meta on at a meta level all those sorry sorry to interrupt ask so because what you mentioned it was it was quite quite interesting because that that Peter levels examples and and the the other samples um you they spend like um you know a lot of PhD python code to do it so I really want to emphasize that you could theoretically and practically that something that we want to show people can create these Bots using Nexus GPT right like using no code if they have no idea like how to code in PHP python or anything whatsoever exactly and and that's kind of what what we'll try to show uh in in this demo is really kind of you know taking like even the most simplistic one or a more complex one and showing you step by step within a few minute how you can create those agents without coding or even without knowing how to prop prompts perfect that's what what I was looking forward in in the workshop right um they they have this like gazillion years of experience um to like code where we can make these similar things using no code um you can make it fast you probably cannot make as much money as they are right now you could um they've been spending a lot of time in marketing followers and stuff but the development part can be cut down to like virtually minutes um as as as as we're going to see super soon I just wanted to emphasize that sorry to interrupt go ahead please no no of course and and I think you're you're touching a very important part which is actually like the most difficult thing when you're building something is not really about building the product uh and you should spend a bit of time building the product first and having kind of a vision but most of the time should be spent on sales and marketing uh it might be counterintuitive to say that but you know having something that's thick and that people are willing to buy uh takes a bit of effort and takes a bit of iteration right so you need to have something that you can iterate on faster right so if you take for example the example of P levels with therapist AI it's a bit of a skewed example because Peter has a lot of followings right and he has access to a lot of marketing channels that enables him to grow things very fast right and to quickly iterate on the marketing part right so for him what's difficult is not really the marketing or getting his idea out there it's really the building right you have kind of the problem if you're a new entrepreneur is that like the building should be the least difficult part in your journey but like actually selling an AR marketing is I would I would say the most difficult part 100% couldn't agree more like like sale sell first and then build it I feel like that's the new Norm right now and something we're emphasizing in the hackathon as well from day Zero if you don't have the product at all you have you just formed a team start trying to find that user um Alex had a question um if we can get get into it super quick will getting into startups as a no code Builder limit yourself in terms of growth and scalability for the startup so can you repeat the question so will getting into startups as a no code Builder limit yourselves in term of growth and scalability essentially asking if you use no code do you do you face scalability issues in the future yeah I I think yes and no right if you if you use no code it's it can be an issue long term if you scale right but there is such a thing as technical depth and basically what technical debt is is that you're saying hey essentially I'm okay to make uh um to to do something more simple right now and see if that works and if that works and if people are willing to buy and people are are building and buy buying a lot of things from me I can afterwards refactor my infrastructure and go and make something a bit more reliable so obviously you know the the biggest hurdle and biggest issue in your scale journey is not really the product and the code because those are the things that evolve as your startup grows but it's really about like you know getting something that stick I would say I don't know if that answer your question I think so Alex if it did answer your question put a yes or thumbs up or one in the chat yeah please continue if something else comes up I'll I'll interrupt you in the middle so so I think uh basically right I I wanted to kind of go into different categories but we don't really need to go there but basically this is kind of a a very high level kind of categorization of all those use cases right from being able to do semantic search analyzing document ideation categorization content generation those are all the things which is kind of a framework to think about like if you want to use llms and AI whether it's agent or not probably your startup and your ID might fall into one of those eight categories right and so it's a good way of like thinking through those categories to generate IDs right you say hey how can I make something that for example analyze document for that type of job right what type of document analysis does a real estate broker needs what type of and you can do that for any job like what type of content generation does a real estate broker need what type of text categorization etc etc so you can use that as a framework to kind of s like come up with idas of agent and then test out these uh these IDs in them um but now like the the big question before we go into that is like what actually is an AI agent right um and so the thing is when you look at an AI agent right the actual AI which is the large language model is only a part of that technology stag that you need for an AI agent so an AI agent in essence is an AI model that is able to autonomously perform task on your behalf and potentially is able to plan around those tasks to achieve broader objectives that's kind of the simple version of it and the synthetic version of it but what I mean by that is that let's assume that you give an llm like you know Chad GPT a bunch of tools that are proprietary to you like oh you give it like a set of knowledge which is all your product you're you're giving CAD GPT also uh access to your email address and you're giving Chad GPT ability to write specific things that you trained it on right now you given that that thing to Chad GPT and now you're asking Chad GPT to perform an action which is for example hey can you look up that website write an email and send it uh to that person so that you know they can be aware of my product right what shbt will do is probably perform singular task or series of tasks one after the other what an agent will be able to do if you give an llm all those things uh what agent is able to do is to understand that your objective is to sell your product and then look at all the tools and the tasks that it has at its disposal to arrange them to make a plan which is hey in order to do that first I need to look up that website of that lead then I need to write an email in the style that uh I've been taught and then I need to use the tool to send an email through the email address I've been given to send the email to the person and that's kind of kind of what an AI agent does it's able to kind of Reason in a sense around an objective to be able to perform it um obviously right now if I'm being frank the the large language model are not as stable enough to be able to consistently do that but that's probably something we're going to see a rise on in the next couple of months to a couple of years is really kind of starting these emerging agentic Behavior as we call them which are kind of this autonomy of those agents in the in time you can still be you you are still able to perform those task but you'll have to kind ofing be able to hard code part of it and say hey when I'm asking you to do this you do you you always do a and c and you kind of teach the agent as it goes to perform series of task for you um obviously you know as you said uh before the stack of an agent is really difficult there's a lot of coding involved and that's kind of in essence why we built uh Nexus right is to be able to be an abstraction layer of all these things and to be able to build these agents uh without touching a line of code essentially um if you don't have if nobody has more question what I'd like to do right now if it's okay with you is to go into demoing what I mean by AI agent and what I mean by building an agent absolutely just a quick uh quick question from atolia thank you so much for joining it toia can you touch on how expensive are these agents to maintain yes so maybe you can sorry give like an example of an agent cost anything anything um you can share so it's it's it's very it's really a case-by case basis right so how expensive they are is essentially how frequent they they are used right so you have to think that the main engine behind an agent is a large language model right so depending on the type of large language model you have different rates per tokens right and depending on how much Freedom you give to that agent he's going to call that large language Model iteratively A lot right so potentially you can go from if you go from a very simplistic uh large language model that cost very little per uh token uh dollar per million token Al the N has limited ability all the way to a large linguage model that a that that is very expensive and an agent that performs a lot of task in it it goes from a factor one to a thousand to be honest so there is no one size fit all answer to your question sadly it's really about like being able to test and iterate on it one thing though is that what a lot of people are doing when building these agents is to be able to more to in order to do more cost control what they do and what for example Peter levels did is that he hosted his own large language model internally so what that means is that he took an open source model like Lama 3 he made it run and he say my agent is now running exclusively on that Lama 3 and what that means is that he had a cost of like a dollar an hour uh or a bit less than that to run this Lama 3 and he run it indefinitely and basically all the requests went from that so his cost was capped to a dollar an hour essentially this makes sense when you have enough uh request to your agents if you don't have enough user it doesn't really make sense to do that yeah super valid point uh quick touch upon that aulia it's like asking how much does it cost to run a SAS or a startup the the the answer isn't as clear as uh us mentioned and second thing there's like a bunch of these token calculators that you can use online and if you just like try it you'll actually realize the cost and just ask more people like awesome and other people who are running agents to see how much it costs thirdly um again the same thing once you hit a critical point a the cost you're spending um doesn't matter because you're making the revenue be you could always lower it c um once you have enough data using gpt3 GPT 4 maybe GPT 5 you can use that to train a local model like llama and that could potentially reduce your cost as well just wanted to add that in parenthesis as well that's a very valid point and that's a lot of for for very specialized use cases uh it's what a lot of people are doing right so they they generate enough data on a higher model and they train a smaller model that is specialized um frankly you you might see less and less of those because you have those open source model very small open source model that are as capable as larger larger models so to give you kind of an idea uh Microsoft released 53 uh a few weeks ago and those are not you know to give you an idea in terms of how many parameters uh the big Lama 3 Model that we've seen uh so far has 240 billion parameters and like the big Lama 2 model had 70 billion parameters model right now what Microsoft was able to achieve in terms of reasoning capability with 53 um was on par with Lama 2 if I'm not mistaken with a 4B model right so you had like something that was as efficient as something that was 12 to 15 times bigger in terms of computation capability right so over time those reasoning capability are being commoditized it's going to cost less and less to run them perfect thank you awesome yeah super valuable and in quick side note again aolia I think has an AI newsletter right now we is planning to launch one so feel free to give Nexus GPT a shout out and um and check out this tool with the and and get this tool out to other people um do you want to do the demo um we're we're super ready for it yeah yeah cool so basically I'm on Nexus so if you go on gp. Nexus that's how it works uh you sign up you ask a few qu you answer a few question and you're basically ready to go so we have a kind of a very simple build your own agent type of things on our website so if you go to the home basically what you do yeah sorry to interrupt so it was super fast I got lost a little bit um very very first step what website should they should they go to so on gp. Nexus right so if you go on gp. Nexus you basically on our homepage and here you can even see like a few use cases of agent from customer support marketing Etc if you need some inspiration you can see them here um and on chat with your agent you you need to sign up to it uh and then basically you have a quick onboarding of basically creating your first AI agent and integrated that AI agent um and those are kind of like basically what what you have to to to go up and running on Nexus once you are there um you you're going to click on create your first agent right and then what you're going to do is that that you're gonna you're gonna basically specify everything that has to do with that agent right so whether it's a specific role or a specific capability or going into working for a specific company you can specify everything there like for example if I say forya l who is like a let's say a customer support specialist and at boud which is uh startup accelerator right and what do they s sorry could I be a community specialist um my title okay Community specialist um what does your company sell for you what what do what does the below good question I feel like I should I should know that clearly um it would be um education um community and uh fund as well so Capital well what we call the three C's content Community Capital that's the one yes all right or people you will be interacting with is startup Founders I'm guessing yes Founders creators and Builders yeah creators and Builders what's the main surface of that agent uh providing supports and mentorship does that sounds good to you yes all right what should I do if I cannot find the answer so send an email yes or or Discord we do have a support ticket some people keep pressing it randomly just please don't yeah what's the what's the link to the Discord it's B loud be l o .xyz Discord all right what's the writing style of borea oh it would be it would be uh chill fun friendly um uh explained like five type of thing yeah all right and basically like that's what uh we we need to create like a a good first agent right once you're there right what Nexus is going to do is based on all the specification you provided on that agent it's going to give you ideas of task that agent can do right so it's a community specialist right so it gives me kind of ideas of task like content creation event planning Community engagement Mentor matching feedback collection right those are kind of task ideas of task that agent can do like hey creating written blog post to educate startup planning event Etc you can create your own task obviously and task essentially what a task is you have to see it like a a GPT task so it it's using an llm to perform a very specific type of writing or or creation or transformation of document that your agent can use right so if I say hey write an email it's going to use the task email in order to write a specific email the reason we did that is so that you can separately and independently train this task and change and iterate on the prompt of that task to go into creating the content that you really want and really need right but let's assume I'm all good with the suggestion and and basically once you've done that it's creating your agent and you have your first agent but you have nothing to do it's going to out of the Bosque it's gonna it's going to create all the prompt of that agent as well as all the prompt of the tasks uh of that agent right so so you don't have to kind of be a master prompter we we're we have like an internal Model H pre-trained model that do that for you and it generally take a few seconds and it's going to go into all different task and that's that should be about it right you have Poria and I can go in before going into adding knowledge I can go into Poria which should be around here that's Poria and you can see here that it created The Prompt of pora right so it's a community specialist here's the role you you do that here's like how you communicate and those are the rules you follow when you communicate and this is kind of an idea of like how the should behave essentially but if you go into skills you have all those skills that are being created alongside it which are which has their own prompt like you know cont content creation it's like hey as a Content specialist this is how you do and those are the information you receive to create the content and those are the type of content you do right and obviously you can iterate on those prompts but that will give you kind of an idea of like start a prompt for that agent and from that Point your agent is ready to go you can just communicate with it um but we can go even beyond that if we go into integrating specific knowledge of that agent right and what I mean by that is that you can add document Notions website or anything of that agent to make the answer more more relevant uh to that specific agent right so if I go into for example an in url um if I go into beloud doxyz right I can it's going to scrap basically the website look into specific sit map of that website in order to find all the pages linked to that specific website right we uh sorry so we might have enabled the robot um no scraper thing I'm not sure if it'll work but it'll be interesting to see if if it works yeah yeah yeah so far this is it's a it's like a a Wonderland right um we we tried to create create an agent like nine months ago 12 months ago and it was a pain of like trying to just set up the basics and this one just like set up in three minutes which is pretty cool okay it seems like it's it's Gathering the the links yeah and you have all the links in all the data and basically you click on continue and it's gonna upload the data to your agents and it's gonna not only upload the data but if you go back into uh pora right what it did essentially on Poria let me put it back up is that on the specific prompt of porya right it added a part which is additional context from that knowledge set right and that knowledge set is basically will be trained on the website so that every question porya receive it's going to find on boud website what is the most relevant snippet of text to answer that question and it's going to answer that question from knowledge you have on your website essentially that's so cool um so sorry a question that comes to mind and if someone else has any questions or thoughts put it put in the comments um what is like the the context window of the prompt like how how much is it because if it's also having the prompt and also the the knowledge from there I feel like it will it will increase a lot so it really depends so the way we we do by default the chunking it's really it has to do with how you chunk basically the text so what we do by default on Nexus is that the chunking strategy you can use is what we call the how do you call that the semantic a semantic chunking strategy and what I mean by that is that it's going to look and read the text and it's going to find break natural break point in the text to create small chunk size right so that you have like a few lines of snippet of text on on top of each other and it's going to limit it to a maximum of three to for snippet of text for every question right not only that but when we create and when you add knowledge to the agent what we do behind the scene that's why it may take a bit of a bit of time to add the knowledge is that it's going to look at all the chunk of SE of text and it's going to Cluster uh chunk with similar uh content together to create summaries of those chunks or or those clusters right and it's going to do that it until you have kind of a tree structure or a pyramid of your whole knowledge right so that you have not only very abstract type of knowledge but also very detailed and concrete type of knowledge at the same time gotcha yeah that that makes kind of more sense now but super excited to see if if I could actually answer the questions to the points I'm not gonna lie to you it's it made you have 150 pages on your website so it's not going to be in that demo it's not going to be done it takes like a 20 30 minute to properly scrap and and clusterize everything uh obviously if you have like one document it's very easy but right now we we i' asked it a very big task yeah no no worries yeah it's it's probably our our fault we created too many pages there's G to be so so many more too as well we do want to create an AI agent to create more pages as well oh well you could do that with Nexus right yeah oh we're definitely gonna try try it after this once you have that AI agent right one of the cool thing you can do is give me one second so I can show it to you so so what you can do is that you can integrate that agent into different workflows right so if you look at integration right now we have like different type of integration from slack teams embedding it directly into your website a telegram WhatsApp SMS Facebook messenger and and basically what it means is that you're making your agent available to the world so if we take for example the example of picture levels he had an AI therapist and he integrated it on telegram right so you could have porya if I can find Poria somewhere where is Poria here he is yep and basically you have Poria Telegram and telegram integration and basically you need to have a token from what on telegram from a botfa right so basically you go into your botfa you get the token and then once you have that token you can just add it let me try to create a new token so it's going to be easier new bot yeah and sorry just a quick comment it's yeah BFA is super easy and uh it's it's good that you put the instructions there as well um because um some people might not know how to get the API token so that's super helpful BFA is super simple you go you click on it the the the the the Slash new all the instructions are there takes like two minutes and the good thing about telegram is that you don't have to worry about creating a UI so you have the chat bot you have the infrastructure of telegram you can just like optimize your prompt your knowledge base uh make it look good and try to focus on like finding the user the marketing the selling so your time to publishing would literally be cut down because you don't need designers anymore right and and sorry we do have this joining the hackathon but you are useful we do want want you to be there but just in this specific scenario if you're using telegram you don't need to worry about UI yeah exactly and and and it's more kind of a tool to quickly iterate on whether or not something is valuable right to give you an ID the first version of Nexus was a telegram B uh and that's how we realized that there was like uh some kind of uh need for that right so no no way so the first version of Nexus was a telegram bought yeah yeah now you can now you let PE other people create telegram Bots that's that's pretty cool and so basically once you've done that uh that's it you have your you have your agent it's on telegram if I go into share your Bot I have the link and you can start chatting with that link right so let me I don't know if I can see you're not gonna be able to see my whole screen but yeah it's here it's SL do you mind putting the link in the chat so I can also send it to people they also try it live as well well yeah and while we're doing this I think um mnaz had a question what are the ways that what are the ways that users can interact with Poria the bot I'm assuming when you showed the Integrations it answered their their question yeah yeah so you have by the way here you can add a payall to Poria right so you can make money from from Poria right so if you say hey oh no way I want to be paid $20 a month uh for Poria and you give them a prompt message and you say hey how many free message am I giving to people uh before porya is paid right so here for example I say hey I'm going to limit to F uh three messages uh second up $20 and I'm going to limit it to three messages and please pay for below um and basically what it does is that it's going to allow three messages and then it's going to ask to pay um and that's kind of an efficient way to start like running a business I would say or like testing an ID uh and then after basically your money is directly on Nexxus so you can retrieve it we're not charging anything on top of stripe fees uh for people generating money it's really about like you being able to build whatever you want and whatever you feel like building want you to do that so sorry so you said you're not charging any fees on top of on top of the Stripe Right because usually platforms there like a 2 3 5 10% markup you're not charging any anything at all no no so so stripe is going to charge you I think something like three a few cents and a few percentage and we're just charging you back that when you when you retrieve your money essentially crazy that's that's pretty awesome so you could essentially make the agent in like few clicks um fully no code um connect it to telegram this simple um i' I've used the telegram API before it's it's not as easy as like just visually so you have to go through all the codes so this is this is quite simple and the zero fee I love that part because it makes it more accessible to people and it just wants people to like build more stuff try out more ideas and you know if they end up making money with it um fantastic yeah and by the way uh here you know we can also test you have a playground where you can test the agent hey hey Poria how are you um and normally porya should be live and answering you back exactly uh and you can ask porya to perform things right so hey can you do some internet uh searches on AI agent and then um create for me some content about it um and so basically what an AI agent is is going to do right is is going to try to assess like the complexity uh of a task please do that it's going to assess the complexity of a task and then um based on the complexity it's going to say hey I need to use content creation for example right and it's going to use the content creation tool and and then you know just create the content that you want right so that would be kind of a Content creation on it but you can do something a bit more complex obviously you can ask it to do multiple steps and then it's going to say hey it's a more complex kind of thing I'm going to make a plan that involves content creation but as a last step uh and and do all the research and do everything and then create the content okay crazy crazy uh question could it so so if we ask it to create like a like a tweet for example like turn this into a tweet and then can we ask it to publish it um as a tweet that's a very good question so in order to do that right what porya needs is the ability to add a tool to it right so it needs to has access to the tw Twitter API right so basically there's two way of adding those tools so I I told you about tasks and then you have like plugins themselves right so we have like a preset Marketplace of plugins a lot of them are Scrappers but if you don't find like the plugins you want you can always like if you know Twitter has their own API you could integrate any API you want and have actions that your agent can perform on your behalf so you can add your Twitter account your token and all of that and say to Pora hey publish that tweet and it's going to be able to do that that's awesome the the reason I'm asking is one of the people um on this word NAIA they had this idea so that you would create um an event and then publish it so a telegram B that lets you create events and then it would publish it on the event management platforms so I just wanted to check in to see like you can add in custom apis as well yeah of course you can add so to answer kind of your question the way you can integrate your agent is not only through these integration and one of them is like an embed chatbot so you can embed a chatbot into your website into any website and you have like a a link to it right you have a public link to the chatbot so you don't have to do anything either uh and you have an API to Nexus right so you could use the API to Nexus to build uh and to integrate the chat like Nexus as a as an infrastructure for your Bot in your prompt so you can kind of manage your agent manage its Tool uh the different plugins the task The Prompt Etc and then just connect it to your to your app and you do the front end if you want to do something a bit more complex you do the front end and then like the whole back end can be Nexus good that's awesome uh Ali had a comment imagine an agent for hacking other agents as well um I'm assuming so we have the we have the tasks that the agent can identify to do can agents interact with themselves can a task be an agent itself that's a so it's a very good question so multi-agent is we feel not yet super stable to properly do and we don't really want to have agent like we we feel like an as a practical concrete use case that is you know reliable uh we didn't feel like multi-agent could be useful as an automatic mode however when you are on the playground right and let's say you can always kind of tag an agent and say hey porya how are you and then you can within the same conversation you can tag another agent to pass the ball between agents and say Hey you can and you can pass to like a French agent and and the French agent is like answering back right so oh nice it speaks better French than than I do um so another quick question um might might be possible might not be because this is on the web it's embedded within the infro so you have total control over it because that's one of the other benefits of having your own website platform you know the front and backend but if you use telegram or Whatsapp or slack can you use this tag in agent functionality to there as well so say if they're talking to pory on telegram can they tag like um awesome from from at the time no because when you're integrating an agent uh into our integration you're integrating one single agent you're not integrating Nexus per se right so you you you're you can integrate like one specific bot on telegram but you can't have really kind of accessibility to all the different Bots because this is kind of more something you can do on Nexus gotcha um Alex has a uh wait Zach has a question how about adding to web codes GitHub what what do you mean by that um um yeah uh you I mean basically to for all those tools and plugins to be frank with you like the world is kind of your oyster right here so as long as it has an API it can be integrated perfect and Alex is saying how do you implement an AI agent into an actual website instead of just implementing into something like telegram so here you know we have our user Gra and you have like the API references I would say that's probably kind of the easiest thing you can do is to have your API key for your agent and just start like integrating it into like very specific concrete use cases of specific websites so I would say just use the API perfect because you also mentioned you have like an embed chatbot feature as well so I feel like yeah that could also be another way and if we could also send like curl requests like post and get um I feel like you can integrate with zap here and and make all those no code platforms too as well yeah as well yeah it's true okay perfect sounds good and like the last thing is like the model right as I told you you can swap the models uh have your agent run on a different model the task on different models as well you can host your own model on Nexus if you want so if you want like Lama 3 you can select it run it and it's and deploy it and that's it um it's like two clicks as well perfect yeah every I feel like everything with Nexus is two clicks um probably you could change the name to two clicks um as well just maybe a branding thing to consider um super amazing so how to create an AI agent in five minutes or 55 minutes using no code but we did go through like the whole thing um quite interesting um I didn't know you could do the tagging agents that was quite quite nice live I saw live data um I didn't see you touching upon that how how does that work because you can upload like you know CSV notion I'm but what is live data so live data is essentially data that changes right so if you look at knowledge right you you have like right now it's a notion right but basically the goal here is to have like any kind of database any kind of Google drive or anything with Doc or like Google spreadsheet that changes and that change dynamically uh the idea is to be able to kind of connect those data directly to your agents so that they can perform tasks that are up to date right right so one of the thing we're thinking of uh of adding is just like a being able to connect a database right so if you have like a a super simple database that you that you've built you should be able to kind of connect that agent to your database so that it always has the latest data that you've had right now you can connect to notion obviously but you'll be able to kind of connect to more use cases in the future that's awesome a quick ask I know you're doing some updates internally as well but uh we have couchbase as one of the other partners for the hackathon um they have a database platform they're a database um can you possibly in two days or three days add them as an integration as well so if people want to use couchbase as database they could get live data from there so in real time people could like see what's happening so if we have some people from couch base that are willing to uh help us uh through that integration on data yeah happy to to look at it with them perfect sounds good yeah have like a gazillion team members I'm most likely sure they could help out but I have to reach out to them as well that could be an interesting case study as well somebody else also had a question and for folks staying um on welcome Sam for those staying on we do have two more workshops after this one with repet and ICP stay tuned um mnaz says what is the difference between an agent and an AI chatbot yeah so so I tried like touching upon upon that during my presentation but essentially the way and I'm going to go back into myself so the the way you can see that is an AI chatbot is or an AI assistant is basically something where you ask it you give it task and you say please write this please do this and it's going to perform the task an AI agent has ability to perform those tasks with more ambiguity so you can give the agent objective and it's going to find a way to achieve the objective Without You NE necessarily specifying the task or specifying the series of task it needs to do so you have to think like a a chatbot is uh an intern that you ask it to do specific things and you ask like very specific instructions and then agent is like a a young employee that just started that has kind of some kind of knowledge of what it can do and kind of do and can like arrange the stuff itself to achieve an objective God something I might add and correct me if I'm wrong um awesome so we could also have like an AI agent um AI thingy and the the output of it the representation how we interact with that AI agent could be in the form of a chatbot it could be in the form of of I don't know what other forms are there besides a chatbot is so the the thing is whenever like an AI can be just an in and out right so if you if you say like I have a piece of a field on my website and that field is able to produce something right whether it's like another like we we've seen like veral creating generative UI in their in their in their bot and all that this is essentially like a field that you ask the weather and it's going to give you back like the weather on a specific location with its own UI right but in the background what you have is some kind of agent who is able to kind of retrieve the right piece of UI and plug it into that call in order produce like the the good weather app so like the agent itself and you're right I think pora the chatbot is the interface while the agent itself is the process to totally uh somebody said yeah software so chatbot uh as an interface interface software on the search it could be just like behind the scenes right um that sounds great any any if there are any other questions um please put them in the chat anything else you want to show us um from Nexus um awesome any cool feature or any any ideas like because people are going to be using Nexus um during the hackathon some of them most of them um it's so cool I do want to use it uh what what do you want to see built using Nexus I think one of the cool thing you could you could really do is think about like a specific type of very simple type of coaching or any type of you know conversational uh AI that is trained on data or knowledge that you only have access to so if if you can do something that nobody can if you know things that very few people know and your knowledge itself is valuable I think you can create your own kind of version of yourself very easily and just upload that into your agent and be able to interact with it and I think if you do that as a first step it it's really good because it can give you a feel on uh how to do prompts it can give you a feel on how to iterate on things it can give you a feel on like what's the right model to Pro perform these actions it can give you a feel on a lot of things and like and you you'll be able to kind of just quickly do the whole thing right yeah s sounds good uh Leon asks do you only use models from open AI or are there other models you like no so virtually any model right so we have like by default uh they run on GPT 4 Turbo but you can make them run on any model of op any model of clo any model of mistl or any model whatso ever if you can connect to it uh if you can even run your own model locally and connect it to Nexus uh and you can decide on a model and run it in the cloud and deploy it on our gpus and run that agent on it right so virtually anything perfect and Alex oh there is an echo okay Alex says do you have any tips on getting startup ideas or ideas for this hackathon that's something we'll share with you after this that you can build with Nexus GPT so um based on what I understood you said um look look at yourself your team your your expertise see what you know that other people don't and there are a lot of Industries out there so don't be limited to your circle so go out Branch out like uh medic The Med medical area the agriculture Aviation think of Industries older industries that are using boring tools that you can also disrupt with a simple chatbot agent you know Peter levels as you mentioned he went after therapist so you go after hairdressers go after uh moms or or veterans pick pick one of those niches target audience to to address 100% And and you know the the thing is uh one cool thing that I forgot to mention is that when you upload documents to an agent those are you can enable multimodality on these documents so it can like the agent is able to understand like text images and stuff like this I'm telling you this because one of the user is is making chatbot of specific instruction manual on Industries right so you people can just upload uh machines manual very complex machine and you have a trainer shat bot like that's kind of one use case right another and you have like those use case on a lot of different industries that that sounds pretty cool I missed I missed the part that you were going to do to demo I had a call and I really want to go back and watch the recorded video but uh can we also can an Nexus go through video to text or so right now um right now like the agent itself cannot perform uh image text or video text or do the other part which is like video to text uh video to image ET but this is some kind of something we we have in in our road map I I think the the text to video and video to text uh is going to be I I think we it's the one that we receive the least request on uh uh text to image image to text was is something that is very near uh basically because it it helps you kind of attack like things that are a bit more creative with those agent and being able to kind of create because when you create an image you also have a prompt for that image like a wrapper prompt of that image and being able to kind of consistently create very specific images can be in some cases very powerful so yeah definitely in our road map but not right now thank you so somebody also asked uh can Nexus do voice detex so assuming no right now yeah exactly one thing I can tell you is that next week if everything goes well we're going to integrate uh phone calls into our agents right so being able to give them phone numbers uh which is good because you know you have a lot of Industry that need customer support right so if you think of any AI agent that can do voice in order to answer questions well next week you again few clicks you'll be able to attach the phone number to an agent and just that agent can be called that would be great that that sounds amazing I really want to uh use that phone call agent that you can talk with like a therapist but you actually can talk to it instead of just texting that would be fantastic also we know that you're chipping in for the price for hackathon you're offering 2K credits worth of credits um and can you let bit a little more about the requirements uh what should what are the requirements that people and participant should should have to beti eligible yeah so for basically essentially for the duration of the hackathon uh we would like to kind of reward the people who are able to make the most money with uh agent right so the people who are able to generate the the most money with the agent will receive uh not only a one-ear pro plan on Nexus so uh on that plus a lot of credit on it so that they can basically their whole infrastructure cost for the full full year will be almost taken care of so they can start like making as much money as possible with this agent in scale of business that would be on top that also offering to the participants like a month of a hobby plan on Nexus so they can actually get going with during the duration of the hack wow that's amazing thank you so much just I just wanted to emphasize that for everyone who's participating and participating the hackathon to have that in mind now that you're thinking about creating AI agents um thank you for being here awesome uh it was great having you here if there's no more questions uh I think we can kind of end the call we have one more question Alex is asking how did you get the idea of creating NEX gbt and why did you decide to make it easy and mostly free to start and create stuff uh sure so uh before Nexus I had an automated marketing agency so basically you know instead of you would go into a marketing agency but inste of people you would have a version of gpt2 then gpt3 kind of answering you uh and we did that uh with associat and we we scaled that quite a bit and then when Chad GPT came out we saw I saw that there was a lot of people who wanted to do the same so I I figure like I have some knowledge on it might as well help people Empower people to do what I did in scale I did um with their own IDE of business that was amazing thank you I'm so inspiring uh we can maybe later on have you to like go through the story of building a startup and how is that going how how hard how easy it was to start Nexus gbt that can be a conversation for next time but if there's no other question this call yeah just a quick note of what what I've noticed is that um most most Founders go through like the Agency Route first uh to make sure that they have enough capital on hand they develop the connections they actually understand like providing the service manually or via AI the AA AAA you mentioned and then they transition to like assassass because they sell first they learn optimize that skill then after that they can automate it and like give people the tools that the shoves on on how to do it so great great job in making the transition um awesome thank you and and yeah I think like you know I think when it comes to actually generating money going into the service R is probably the quickest way you can do that is basically you just have to pick up your phone call and and just start selling to people and you can generate money just like this um and then you know you like the IDE is to be able to kind of quickly productize that into uh something that larger amount of people can can can e but I I think in any case whether you're building a product or whether you're building an agency at the end of the day the first initial customer are going to be the same uh you're going to spend a lot of time talking to them you're going to spend a lot of time understanding their needs and whether you're building an agency or a product it's going to be basically the same Journey at the beginning at least totally makes sense thank you so much for your time awesome I know you have to run uh we're super busy too uh we're gonna have another Workshop in 20 five minutes is so stay tuned uh I think on the day for Saturday we'll have you for uh some some time Lots if people have any questions about Nexus um they can either book The Call with you or like you can be there for like an office hour depending on when you're free so course I'll I'll put a calendly and you know people can book uh 15 minutes sounds good I'm super excited to see what great things people build with Nexus because you can literally spend five minutes to create it link your stripe and the whole time you can focus on on trying to find the user so our our first goal was like finding a user from outside of the hackathon that would try it for free and like get feedback on it then I chatted with you you said we can make turn that goal into a paying user so see who makes the most money in three days somebody else's goal was getting acquired in in the three days so it's going to be a fun one yeah thank you so much awesome you so much this this link is available on YouTube LinkedIn Twitter U we'll also send it to the participants and have a good rest of your day take care take care bye-bye bye have a good day thank you so much for being here everyone uh we have our next Workshop in less than 20 minutes with the repet and the one after that with ICP they both are very interesting I was just talking to replate and they also chip in um one your subscription um as a prize for a team for all of their members repet and then after that we'll be having ICP and ICP is also providing um some cash prices so it's very interesting and I re really much recommend to stay on to hear what they're going to say and to see how we can use our tools and have a great day yeah before before we head off um thank you so much for all those who stayed onaz Alex sh um Ali Sam thank you for coming um let us know what what other workshops you want to see from us're We're super busy right now with the hackathon but we'll try to um give you the awesome experience as well you take her and have a good hi hi everyone

Channel: Beeloud

Views: 1,627

Keywords:

Id: XEPgrrf19d8

Length: 66min 37sec (3997 seconds)

Published: Fri May 10 2024