I found this STUNNING Local Perplexity CLONE!!!

Key Takeaways

I learned how to build a local perplexity clone this is a completely open-source project and in this video I'm going to show you how you can run this on your local machine first of all if you're not familiar with perplexity perplexity is one of the hottest AI startups in the world it is revolutionizing how search could be used while it is powered by llms in the back end the primary stuff is how perplexity has manage to connect all different things like knowledge internet and current events so what we are going to do in this video is we going to use an open-source project which very surprisingly is a go based project primarily go and this project L local search L local search and we're going to use this project to do our local perplexity clone it may not look as sophisticated as perplexity this basically serves the purpose and if you are a hacker like you have got the hacker mindset then this is something that you can download it yourself and then play with it you can build things on top of it and that is exactly why I decided to make this video first of all I would like to show you a quick demo so I've got something here so I can refresh this once I refresh this it gives me a bunch of options and I can just go ahead and then ask some question okay how much does obsidian sync cost now what it does is when you search first of all it takes the query from you it starts looking for your local information the vector DB that you have got in this case it uses chroma DB and then it looks for whether whether this query is available before and also it starts looking into internet and then it starts getting that information acquire that information store that information and then augment that information with an llm and then it gives you the answer and what is the llm that we using in this particular case we are using hermis to Pro mistal and for embedding we are using all mini LM so as you can see here it successfully has extracted the information so the cost of obsidian think is $5 per month I'm not sure I'm not validating or verifying this information but what it does is it gives me the sources from where it got the information from Reddit hacken news obsidian form Forum the tech Republic which is I think a tech blog so it gives me all the information that is required for me to understand whether the given information is valid or not so I use my knowledge to finally validate whether this Authority that I've got here is good enough for me to verify this information and this is kind of what perplexity actually does while perplexity is much more than that uh it at the fundamental of perplexity when perplexity was started this is what basically it was kind of doing you get an answer for the question and it doesn't give you only from an llm it also gives you from the internet and it gives you certain referral articles like references which you can use to validate whether the information is true or not this is a startup that is supposedly challenging Google to change how search is used and that one we have got completely local now thanks to Neil herik So now how do we run this what is this L local search does not require any open AI or any other API key so you don't need to use Google Gemini you don't need to use CLA you don't need to use open AI so how does it work all it requires is Ol Lama so you need to download two models on Ama and then you need Docker you need only these four things let me repeat it again you need two models two different models one is hermis 2 Pro mistal it is from newest hermis and the second one is you need all mini LM this is an embering model these are the two models you want and these two models will be run through ol Lama and then you need Docker if you have got these four things on your local machine very first step let's open our Command Prompt so in in if you have if you have got a Mac open your terminal if you have got windows open your command prompt whatever that you have got shell open somewhere where you can download the git repository so I've already got local search here so I'm going to go into a different folder and then maybe like on my desktop I can do this I'm going to go to my desktop and then all I'm going to do is clone this repo so this is the repo where the content is there I'm going to just clone the repo and as you can see here it's very simple and straightforward to clone it I'm going to copy here and then clone is it is going to clone from the main branch so if you're familiar with Git You know that there are multiple branches sometimes available so it's going to clone from the main branch and after it clones from the main branch then all you have to do is go into that particular folder so local search is what you need to go into it and you can also open this in your computer and then you can see here so you can see that this is all available in your local computer at this point we have not downloaded the model all you have got is the front end back end and the engine it runs so you can see Tailwind you can probably see some go files here and there and then you need Docker compost so the first thing is you have cloned the repo right you have cloned the repo so that means the source code is in your local computer the next thing that you need is you need to make make sure that you have got a Docker available so in my case I've got Docker available here and I can probably show you my Docker so make sure you download and install Docker Docker is free so you can download and install Docker so once you download and install Docker the next thing that you need is you need ol Lama so what you need you need ol Lama so let me go here search for you and if you're not familiar I've got multiple tutorials on AMA all you have to do is go download AMA after you download AMA you have to serve run Ama or you can open AMA software and then that will show you that small icon at the top of your toolbar or menu bar that means AMA is running or the other easiest way for you to verify whether AMA is running all you have to do is go say AMA served and then if it says hey I'm already being served or I'm already serving in this particular Port then you are good so we have set up at this point with the local files the local search source code we have installed Docker then we have used oama so now what we have to do is we have to start downloading the models how do you download the models all you have to do is do ol Lama pull and the model name there are two models that we need to download one is the hermis 2 pro model and the second one is the all mini LM model Technically when you download this and then just run it on Docker it should ideally download the model but if you want to make it easier you want to avoid any errors then you can pre-download this model just do o Lama pull and then sorry o Lama pull uh and you can give all mini LM so you can do this yourself or you can just go ahead and then run this and then that will be available so now at this point I hope you have got the llm I hope you have got the embedding model I hope you have got the local source code I hope you have got a Docker and I hope you have got AMA to serve the model so at this point the model is running AMA is running everything is available on 11434 and AMA can access like the docker compose and all the other things now there is only one command that you need to do which is Docker compose up so once you do that Docker compost up and that is going to build this entire thing and give you the front end the app in this particular URL in this particular Port which is 3,000 and that is exactly what I have got here so I've got at 3,000 this particular setup these are all the models that I have from wama not only for this particular project all the models but the models that we are going to use hermis 2 Pro mistal and all mini LM so the two models that we need for this and then after we have got everything done all your have to do is Docker compost up that means everything is going to be set up for example you can see here I've got the chroma DB running and that is something that I did not install separately this repo does everything by itself it's really very well set up but also because it is currently very actively developed you can see the developer is actually very actively developing this so you can see some running changes but most likely at this point it works completely fine for me there are certain flickering issues with the front end but it it works pretty much it works and like I said if you just trying to use a perplexity then you can go use perplexity which is a great product but if you have that hacker mindset you want to do things locally you want to try different things then probably you should download this project and then try this because now you can add extra components to this particular project so now what I'm going to do I'm going to just go here and then ask a simple question let me ask something that happened very recently so if you are a cricket fan you know that there is something called IPL happening in India so let me go ask a very latest result whether it can get me or not who won the CSK sorry CSK versus KKR match this finished just a few hours back so technically this is to tell you whether this model can access internet or this perplexity local perplexity clone can access internet and whether also it gives you meaningful information the answer to this question is CSK has one so it as you can see here it has started looking on the internet whether CSK or who is the match winner and as you can see here so you you can see here that it is making those calls it is making those calls and you can see that it is getting that information it is trying to get that information so it says the winner of the CSK Kare match is not mentioned in the text provided so it means from the resource it has got it could not get the result back because this was just a few hours back so let me now go back and then ask another question which happened yesterday who won the Mi match yesterday I forgot who did they compete with but um I think it is DC let me ask this question and then see Mi I match result yesterday so it's going to search the very interesting part about the search here is that it uses a search engine which is unlike anything else I might make a separate video about it if you are interested but that project also has started trending couple of days back and uh it's very interesting to see how one developer has put together or stitch together this entire thing with amazing features like it's completely local it can run on any hardware that can run AMA and it has also got a like very decent UI interface and you can see Mumbai Indians won the match against delh capitals and it gave me exact information and this is quite amazing I mean the amount of things that you can build on top of this is completely insane and exciting and that is exactly why I wanted to make this video first of all I truly appreciate this project and uh the fact that the developer has released it Apachi 2.0 license you can download it build things on top of it I think they recently added funding here so if you like this project if you use it for commercial purpose I definitely think that you should go support the developer here the developer who has developed it meals but even otherwise I think this is an amazing project the fact that I can set up my own local perplexity clone I think perplexity clone could be an overshot but when perplexity came into the market this is exactly what they had and this is mindblowing for a Onan project and now it's a community project go give a star to the repo should mean a lot to the developer and also if you can find financially support the developer you can financially support but without open aai without any other API key completely local all set up in a couple of lines of code you can run your own local perplexity clone I hope this was exciting for you let me know in the comment section see you in another video Happy prompting