What's going on guys, Josh Pock here, and on this channel we've covered many different AI tools, AI open-source tools, and so many different kinds. But what if we could have one open-source tool that combines a bunch of different AI solutions into one? Well, that's the tool

that I'm going to be showcasing in today's video. It is called Big AGI. With AGI on the cusp of being invented and being in the forefront, we have this tool right here which is a generative AI suite powered by state-of-the-art

models providing advanced AI, AGI functions. It features AI personas, AGI functions, multimodal chats, text-to-image, voice response, streaming code highlighting and execution, PDF import, presets for developers, and much more. You can deploy it on Pre-more in the cloud. They claim to be the most innovative and

productive generative AI suite. I'm going to show you how you can self-host it in the next few minutes. Let's dive right into it. All right guys, so all links as always will be in the description down below. The link to their website as well as the

GitHub repo so you can self-host this for yourself. Um, here's the website, you know, here's kind of a glimpse of what it's going to look like. And they have a Discord channel just to quickly go over some stuff. Tailored AI with all these

different models right here as you can see. Some Mistral, OpenAI, OpenRouter, local AI models which we just covered local AI on this channel. Power for UX, private so chats are stored in your own browser, uh, advanced fun call, AGI mobile UI, voice UI, and then with personas, so

This is really cool. I'll get into that in just a second, and then more as well. All right, so they have some pretty cool tools right here. They have chat, they have call, uh, which we'll dive into in just a second, and they also have Beam, so

you may be wondering what is Beam? Well, as users become more familiar with AI, uh, language models, the limitations of the models can become increasingly apparent and frustrating, leading to a desire for more intelligent and controllable AI interactions. Beam, a groundbreaking chat modality in Big AGI, addresses these

needs by enabling users to easily enable multiple models simultaneously, fuse their best responses, and achieve better output, wiser decisions with lower risk and reduced AI hallucination. So they claim this leads to better brainstorming and idea generation abilities, more informed decisions with more breadth of exploration to dilute

and reduce the incidents of LLM hallucinations, and then higher-stake outputs that far exceed cost trade-offs. When engaging with Beam chat, the user sends a message to Beam using Control+Enter instead of just a single LLM. This is presented in the

diagram below. So we can see you could send off a Beam right here. It could go to OpenAI, Anthropic, uh, Google Gemini, and Mistral all at once, and that's what Beam does. And then step two will be the AI merge. So it combines

responses into a coherent best answer under the user guidance right here. So, fuse responses to user. So pretty interesting. Of course, read more about that feature in right here. I'll leave this link in the description down below as well. They have a full blog post on it.

so there's the documentation right here I'll leave that link down below and there's some advanced features and you know going over different models and such we'll come back to some of this in just a second let's go ahead and talk about getting this set up so the GitHub

repo will be linked down below of course and you're going to go here and you're simply just going to go to installation right here I'll actually leave the direct link down below so this is the installation guide it's very simple if you've installed any sort of GitHub repo

before you're simply just going to get clone the repo and then CD into big AGI I think you guys have seen me do this enough times I'm just going to show you right here but obviously you're going to open up your terminal and do so and then

you're going to run npm install and then npm run dev okay that's going to get the uh instance running on localhost port 3000 oh I'm just noticing that it is using tRPC here so that's cool anyways so once you do that of course you're

going to go to localhost 3000 look like this okay so you will see that we have the chat right here now I'm just going to move this over just so you can see over here on this side we have chat um um and we also have chat so Persona

replies write so append a message draw so AI image generation and then reason plus Rea uh reason plus act so answer questions in multiple steps I'll show you that too in just a second we also have Beam right here so this is what we talked about and then we have call so

quick call regarding this chat so pretty cool stuff here we got the AI Persona so we have like default developer scientist Catalyst executive custom persona Persona Creator so right here it says start here so if you click here it's going to show different examples like help me plan a

trip to Japan what is the meaning of life typical examples you would see prompt right here is going to show you the system prompt so you could customize this you're an assistant right here it says knowledge cut off this is the default one current date Um Mermaid

rendering enabled plant uml rendering enabled SVG and markdown rendering enabled data presentation prefer tables Auto column so you can of course customize this if need be you can also change the titles right here and what are be what AI personas are going to be displayed right here over up here we

have the um LLMs so you can choose an LLM right here I'll show you how to configure these in just a second and then here is the AI personas as well so you could choose default custom whatever the case may be over here we have the

chats so you could search for stuff in your chat you could create a new chat you could delete a chat import chats export chats delete chat and then use folders so you could create new folders like test like so save that you can create another folder like folder okay

think you get the point you could add a chat to a folder like I add this to folder and now we see the little pink folder icon here we go up here to the models and then click on models this will enable us to to configure our models so we could go Anthropic Gemini

open AI, we'll just do open AI for simplicity's sake, and I'm just going to paste in my API key right here. GPT-4, so you can enable different ones, so 001 mini, 001 preview, real-time voice, um, GPT-2, you know, all the different

ones you really want. You can see you can refresh the models. You go all services right, so you got a bunch of different options here. You can also Advanced here, which allows you to put an API endpoint, so you could use Helicone, Cloudflare organization ID, a Helicone key here.

observability, moderation on or off. Okay, once you have that, you can go ahead and actually use your model. So right now we're using GPT-4. I can just go ahead and say, "Hi." You'll see that we have that little gift there chatting, and "Hello, how

can I assist you?" Now you'll see here we have different options right here. If we do different, uh, forward slashes, so we can do forward slash Beam, forward slash draw. All right, you could drop files, um, so let's go ahead and do forward slash Beam, forward slash

beam high, and then we can see we get "Welcome to Beam. Beam mode, so Beam is a chat modality, blah, blah, blah." Explains it, so we can go ahead and do "Next." Okay, it's explaining it here. Go "Next," and then we can go ahead and do "Start." So this is

actually where you can select your different models. So we could do GPT-4, we could do Claude, Sonnet. Okay, so of course, you're going to want multiple LLMs already saved in here, so I'm going to paste in my API key for Anthropic all

right so now I can do GB40 claw son 3.5 now you can easily remove models like so or you can add more models to the beam you could do like eight times four times two times by default you could add you know a bunch like let's go to eight here

yeah so eight is the max I'm just going to do two ryp of pouring thousands of dollars into appointment setters only to watch leads slip away imagine having a team of elite sales agents booking qualified appointments for you around the clock no more wasted time on

training no more frustration with performance and no more draining your budget on inconsistent and expensive call centers introducing stride agents AI powered appointment centers that work 24/7 never get tired and book appointments while you sleep trained on thousands of successful conversations our AI agents outperform human teams

at just one-tenth of the cost join the ranks of businesses that doubled their appointments and booking rates in just a matter of weeks don't get left behind in the AI Revolution visit stride agents.com now and transform your entire sales process with cutting-edge AI

technology it's time to accelerate your stride with AI agents and you can see down here we have our merge model right here so there's fuse guided compare and custom and then we got merge model selection right here I'm going to go ahead and test this generate

we can see generated right here do generate for cloud sonnet okay so it looks like cloud sonnet was a little bit um longer I noticed you wrote high beam which seems like it might be a typo would you like me to blah blah blah okay so now we can see that we have access to

the merge model now so you could select a different merge model, whatever one you want, you could fuse them so a add Fusion AI combines conversation details and ideas into one clear comprehensive answer. There's guided so add checklist, choose between options extracted by AI from the replies

and the model will combine your selections into a single answer. There's compare so analyzes and compares AI responses offering a structured framework to support your response choice or custom to fine tune your own diffusion prompt. So pretty um a bunch of different, you know, varieties that you

can choose from here. We'll just do a basic fusion and of course I just said high so we're going to get some basic stuff um combine response hello it looks like he may have intended to say hi or possibly had a typo with high beam if

you have any questions or need assistance feel free to let me know how I can help. Okay so this one said how can I help, both of them asked let me know how I can help. That's pretty standard um but I think you get the point okay now I'm

going to do another beam just a little uh you know one that has a little bit more um you know content in it and also too you can see right here this is pretty cool, it actually has um something you hover here and it shows you the estimated um so we can see available

tokens, message right here, model max tokens, this message contains about 16 tokens, history so max response right there, chat turn cost, so you can see all the different statistics for you know how the cost is um and you know how much tokens you're using etc. But let's go

ahead and just beam this question right here I'm going to use the same models just click Start here and you can see both of them are generating like imagine if we were using eight different models right here all right so as you can see we got you know a paragraph from each

different LLM right here so we can see this one's they're both talking about the history so let's go ahead just do a basic F Fusion right here so I'm going to fuse them and we're fusing them using gpt-4, but imagine like you could add

for example uh like Grok you could add Mistral um you could add it would be cool if they had like perplexity too um you know just cuz that's really good for Powerhouse for research but anyways and then you could combine them with like any model you use 01 to combine them

right so here's the combined response I'm not going to read it all but you could simply just copy you could use this message so now this is the message that we use from the beam now it's in our chat right so we're getting the best

of both worlds from gpt-4 as well as Claude on it you can use 01 with Claude on it right really get the best response and see what LLM gives you the best response for a certain um question that's a it's a really good

way to compare actually all right if we do forward slash you'll see that we also have draw react browse so we'll assist you uh we'll download the web page assist system clear and help all right let's try draw so drawing an apple give me a a dozen seconds while I draw an

Apple, it's using DALL-E right here, and boom, we got an Apple. So we can uh click on show prompt, and it will show the prompt. So it actually creates a prompt, tier, a vibrantly colored ripe Apple glistening with the morning dew. It's skin

is smooth and reflects the sunlight, creating a beautiful spectrum of red, yellow, and greens. So cool image here. Um, it's actually nice, and you can click open a new tab, of course, you could save this image. It actually does a really good job, um, like look at those. Anyways, I

think you guys probably seen a million AI images, but look at the water on that and everything. Anyways, we can also click on on call, so the call button I showed you right here, we can click on this, and okay, so my browser I'm using uh Zen

browser, uh, which is like Firefox. It doesn't support speech recognition, issue for some reason. So I'm just going to open it over here on my Brave browser. A side note, I just came to the new section here, and this is Big AGI 1.16. It

looks like they're building Big AGI 2.0. You can apply for early access coming this fall. Um, so they're actively developing it, and I really I do like it. Like you know, even just the the me like the gifts here, like the the little

rodent, um, it just kind of, you know, adds some flare to it. I like the the style of this app and uh you know they have some cool features, but like I was saying if we go to voice right here, this is where you can actually call Big AGI. So you can

call different personas, um, so we have the default developer, scientist, catalyst. So if we click on scientist here, we're getting a call from a scientist. Accept, we can go allow mic before I actually do this. I just noticed if you click up here, you can see preferences so

okay, perfect. There's a dark mode, I like this app even more. Um, then change voice, okay. So I already connected my, um, 11 Labs, so you're going to have to add your 11 Labs API key and see I did so, so I got access to this is like my actual

voice clone right here. Um, you can change the gray UI, voice call feedbacks. If you click here and then click on this Cog icon, you'll be able to access, um, the settings here, preferences. So you have chat AI models, all these different options here, appearance, Zen, minimal, clean, text size

page width, full, medium, um, chat AI, Auto diagrams, that's kind of cool. Auto chat title, Auto questioning Labs, so performance, so you can change these things on, uh, screenshot capture, webcam capture, cost messages, um, you know, all these different options here. So really cool that they

have all these features. Without love is like a tree without blossoms. So that's my voice with 11 Labs. So, um, this where you can invoice, you can add your 11 Labs API key, uh, language, mic timeout, speak responses, and then select your assistant voice here. You can see

shortcuts here, draw. So this is where you can change like right now we're using Dolly 3, could also use Pria if you have a Pria API key, and then tools right here. So this is browsing, so you have browsing, um, which you could configure, um, using

Puppeteer right here using a web socket, um I'm not going to cover all that right now, but it does mention that in the docs which I was uh referring to earlier. If you go over here to documentation, you go to browsing, you can see information on

browsing functionality right here as well as React. So, questions answering with reasoning and actions, you can read more about those two quote-unquote advanced features in the docs. And you have Google Search API, so you can do that as well to get some searching functionality with tools. Now, I can

actually chat with these models right now, uh, just because I'm using my mic on my phone right now to record, but as you see it's not picking it up, but you go ahead, test it, talk to some of these models, have your little AI pen pal, um

also too, if you go over here to Personas, Personas is really cool because one, you could just add a bunch of text like paste text from tweets, social media posts, whatever in here and create a Persona using like whatever model you want. And you can go, advanced prompts, you can see

what the prompts are. So, like, Creator system prompt, you can read this. I'm not going to read through it all, but analyze the text, so you can see how they're creating these pro uh Persona prompts just from their prompts, which is pretty cool. This gives you ideas just in general

and you know prompting, um, I always say like you always want to be kind of studying what a good prompt is because you'll never, what, even if it's a prompt in something like this, this could help you if you're coding or doing whatever and seeing how good prompts are built.

cuz uh that's really one of the most important languages you can learn right now is prompt engineering anyway so here you can also do YouTube which is really nice so let's go go ahead to one of my favorite podcasts and we'll just go ahead and Trump's mandate and extinction

level event uh for federal bureaucracy we can just copy this link right here and go back here paste this in okay and we can go ahead it's going to use these different prompts right here we're going to use gpg 40 and click on create it's

analyzing the transcript so generating you can see the bite's going up this may take one to two minutes so we can see it's working input text there there's this kind of thing that happens in biology called evolution and a lot of people think it's

a continuous process blah blah blah blah it's analyzing the transcript here so the speaker in the transcripts presents as an analytic and strategical uh thinker involved in professional field that bios system systematic evaluation and strategic disruption okay I didn't finish it all but oh wait yeah so it it

went through the whole thing it went through all those different prompts uh to create the Persona prompt which is this so you may now copy the text and use it as a custom prompt so you are strategic disruptor name not specified age mid 40s

to 50s which is like you know you know this typical Persona of um Allin podcast occupation political analyst and strategic consultant cultural background Western likely American personality traits analytical strategic objective communication style metaphorical direct yet nuanced engaging motivations uh efficiency and wow so imagine putting in

any video you want if someone may be speaking in the way that you like to speak and getting a Persona that's this is actually very accurate to how this, you know, video was, um, just how they talk and uh, specifically aspirations, systemic reform, uh, fears and concerns

stagnation, potential negative outcomes, core values, resilience, efficiency, Innovation, etc. so this is I think you guys get the point here you can go ahead and copy this prompt you can go ahead go back to your personas here or you could just use this like really anywhere

obviously this is like you could use it in big AGI or you could use this in Claude or wherever the case is um I think you guys get the point here but there's a bunch of cool stuff that you can do in this tool like I said it's a

pretty cool little AI powerhouse here with some cool tools functions, etc. you could definitely do a lot more than what I even showed in this video I just wanted to give you a general sense let your imagination run wild with this but let me know what your thoughts are in

the comments down below guys what cool things are you going to do with this tool have you used it already are you planning on using it what do you think of big AGI other than that guys if you're new to this channel we upload videos all the time on AI, Marketing

sales, business growth, etc. so if you like that type of content you got some about here smash the like button comment down below and subscribe to stay up to date with our uploads thank you so much for the recent 13k 20K on the way also too

Guys, if you haven't already joined our free community, Shri Community.com, I'll leave a link down below to our free Facebook group and Discord channel. And then also, too guys, if you haven't already joined our free community, which I literally just made, Stride AI Community, you are definitely going to want to do this.

I'm putting my all into this community and I'm going to build it with you guys here. I really want to uh interact with you guys, get some give you guys more free value and just build a good cool community here. So I'm going to be making um an exclusive AI resource vault, behind the scenes stuff that I don't even post here on YouTube. All the resources and more are going to be in this AI community.

So if you've ever got any sort of inkling of value from this channel, there it's going to be tenfold in this community and you're going to get to interact with myself as well as other like-minded individuals. All right, so definitely check that out. I'll leave a link down below and in the pinned comment. And then also, too guys, if you run a business, you need help with AI marketing, sales, implementing AI agents, cool callers, call center into your business, running paid ads, whatever the case is, optimizing your systems, uh book a call down below at executive.com/slapp to speak to myself and my team and we can see if it's a fit or not. Anyways, guys, I hope you enjoyed this video.

Have a great weekend, guys. Keep hustling, keep grinding. I'll see you in the next one and accelerate your stride. Take care.

FREE big AGI Generative AI Powerhouse Suite!🤖 Open Source Any LLM AI Agent - Chat, Browsing & More!

Josh Pocock

다양한 AI 솔루션을 결합한 오픈소스 AI 도구인 Big AGI를 소개합니다.

Big AGI는 고급 AI 및 AGI 기능을 갖춘 생성형 AI 제품군으로 설명됩니다.

AI 페르소나, 멀티모달 채팅, 텍스트-이미지 변환, 음성 응답 등의 기능이 있으며, 클라우드의 Pre-more에서 배포할 수 있습니다.

설명에 제공된 셀프 호스팅을 위한 웹사이트 및 GitHub 리포지토리 링크입니다.

다양한 모델의 맞춤형 AI, 비공개 채팅, 고급 기능 등 Big AGI의 기능에 대한 개요입니다.

빅에이지의 빔 기능에 대한 설명

여러 모델을 동시에 사용할 수 있는 채팅 방식인 빔이 도입되었습니다.

Beam은 여러 모델의 응답을 융합하여 더 나은 결과와 환각을 줄임으로써 AI 상호 작용을 개선하는 것을 목표로 합니다.

더 나은 브레인스토밍, 정보에 입각한 의사 결정, LLM 환각 감소 등의 이점이 있다고 주장됩니다.

Beam의 프로세스에는 여러 모델에 메시지를 보내고 사용자 안내에 따라 응답을 병합하는 작업이 포함됩니다.

Beam이 여러 LLM(OpenAI, Anthropic, Google Gemini, Mistral)에 메시지를 보내고 응답을 병합하는 프로세스를 보여주는 다이어그램입니다.

npm 명령 복제 및 실행을 포함하여 GitHub 리포지토리를 사용하여 Big AGI를 설정하는 방법에 대한 지침입니다.

localhost:3000에서 Big AGI 인터페이스에 액세스하여 채팅, 페르소나, 빔 및 통화와 같은 기능을 선보입니다.

커스터마이징 옵션과 LLM 선택 옵션을 포함한 AI 페르소나에 대한 설명입니다.

폴더를 이용한 채팅 생성, 삭제, 가져오기, 내보내기, 정리 등의 채팅 관리 기능에 대한 데모입니다.

다양한 LLM(Anthropic, Gemini, OpenAI)에 대한 API 키를 추가하고 다양한 모델 옵션을 활성화하여 모델을 구성합니다.

다른 기능(빔, 그리기)에 슬래시를 사용합니다.

여러 LLM(GPT-4, 클로드, 소네트)을 선택하고 응답을 병합하여 Beam의 기능을 시연합니다.

병합 모델(융합, 안내, 비교, 사용자 지정)과 응답을 결합하는 기능에 대한 설명입니다.

채팅 상호작용 중 토큰 사용 통계 및 비용 정보를 표시합니다.

'그리기' 기능을 사용하여 DALL-E를 사용하여 사과 이미지를 생성합니다.

Big AGI 2.0 및 얼리 액세스 신청에 대한 언급.

다양한 AI 페르소나와 11 Labs 통합을 통한 음성 통화 기능 데모.

다크 모드, 음성 사용자 지정 및 기타 UI 옵션을 포함한 환경설정 및 설정에 액세스합니다.

11 Labs API 키 통합 및 음성 선택을 포함한 음성 설정에 대한 설명입니다.

브라우징, React, Google 검색 API를 포함한 도구와 해당 구성 옵션에 대한 개요입니다.

텍스트를 입력하고 프롬프트를 생성하여 사용자 지정 페르소나를 만드는 방법을 시연합니다.

YouTube 팟캐스트 녹취록에서 페르소나 만들기.

마무리 발언 및 빅에이지를 더 자세히 살펴보기 위한 행동 촉구.

AI, 마케팅, 영업 및 비즈니스 성장에 관한 채널 및 콘텐츠 홍보.

스트라이드 AI 커뮤니티와 그 혜택에 대한 홍보.