How To Install Quivr 🏹 - Chat With ANY Document from ANYWHERE! (PDF, CSV, PPT, TXT)

this is being described as an AI second brain and it is truly impressive this might be the best chat with your documents project that I've seen so far this is called quiver and you could set it up locally and it accepts almost every type of document or media that you have pictures video PDS csvs PowerPoint documents it works with almost all document types and it runs locally it's completely open source you can use it with chat CPT you can use it with local models on your computer it really does have everything it's a bit hard to set up but I'm going to walk you through the setup step by step then we're going to test it out let's go alright so this is the interface I have it running locally I can upload any files that I want I've already uploaded one so right now I've uploaded the Constitution and if I click view I can find out more information about it then I click over to the chat Tab and I can ask what is this document about the document you provided is a Constitution of the United States so it has access to all of my documents and essentially how you can think about quiver is truly an AI second brain you can throw any and all of your documents into it and then you have a central place in which you can chat with any and all of your documents they also have a fully hosted version if you don't want to go through the work of setting it all up yourself at quiver.app and I'll link that in the description below if I click on over to the explore tab you can see the documents that I've already uploaded you can create different brains so each brain is specific to a set of documents so let's say you want to create one brain for your personal documents and another brain for your work documents you could do that and this is completely production ready so you could have a server running with all of your personal documents all of your work documents ready to chat at any time it is truly impressive so if I click over to the settings tab you can see that I have the different models right here I can have gpt4 I can have local models running I can also have anthropic running you can set the temperature right here you can set the number of tokens and and you can create API keys so you can code on top of quiver I'm so glad I found this project this is the GitHub repo for it about dump all your files into your private generative AI second brain and chat with it using llm's GPT 3.5 for private anthropic vertex Ai and embeddings all right the setup was not straightforward but I will show you how to do that after the demo so here's the video demo for it let me show it to you so here you just click and you can add any documents that you want here's the readme for quiver I click add it adds it as embeddings we're using Super Bass for that then we start chatting with the documents what is quiver and then using any large language model that it supports it will tell you and there it is chatting directly with that document you can also select the models select the temperature and select the number of tokens I can see all of the files that I've uploaded right there alright so we're going to need a few things to get this set up number one you're going to need Docker if you don't already have Docker installed go ahead and install it it's free Docker just containerizes this entire project and makes it super easy to spin up so you don't have to worry about dependencies you're also going to need Super Bass which is a free open source database that we're going to use next open up your terminal the first thing we're going to need to do is clone the repo so to do that let's come up to the top of the repo let's click this green code button and then right here is the URL and we just click the copy button then let's switch back to our terminal and we're going to type git clone we're going to paste that URL in there then we're going to have and and CD quiver so that's basically just two commands we're going to clone the repo and then we're going to CD into it and I'm on a Mac right now but this should work just as easily on windows so then I click enter it close the repo and then switches into it and you can tell we're in there because it says quiver right there next we're going to need Super Bass so go to superbass.com and sign up for a new account it is free then we'll click dashboard and then we're going to create a new project and I'll keep it under my or Org the name is going to be quiver 2 because I already had one going we're going to generate a password we're going to leave everything else the same and then create new project okay now right here we have the API key we have the URL and the JWT secret all of which we're going to need in a moment next switch back to the GitHub repo we're going to need to grab this code right here which copies two of the example m files and creates the actual ones we'll be using so I'm just going to click copy switch back to terminal I'm going to paste those two commands in there and then hit enter next I'm going to open up visual studio code and then in the top left hit file open folder and then open the quiver folder that we just cloned from there click this back end drop down and then the dot m file is right here and this is where we're going to need to change some settings first we're going to grab the Super Bass URL so head back to Super Bass and right here is the project URL go ahead and click copy switch back back to the dot m file and then we're going to place it right here next we're going to need the Super Bass Service key and it starts with ey and that can be found right here API key now I'm going to rotate this API key before publishing this video copy switch back to the dot m file we're going to replace it right here next we're going to need the JWT secret key and to find the JWT secret key come back to Super Bass we're going to go to Project settings and then API and then that's right here so we're going to reveal it and copy it switch back to the dot m file and then we're going to put it right there next we need either an anthropic API key or the open AI API key you don't necessarily need those if you're going to be using your own local model but for today I'm going to be using the openai API key so if you don't already have an open AI account be sure to sign up and then grab your API key and to do that you go to platform.openai.com account slash API Keys Next we're going to create a new key and I'm going to call it quiver I'm going to grab that key I'm going to switch back to the dot m file again I'm going to place it right here and I'll save now if you want to use Google authentication that's when you would need these credentials right here for the cloud project and the applications credentials but we're not going to be using that today we're going to use standard login and password and if you wanted to use a local URL you would do so right here you would set private to true and then you can give the path to your local model so we're done there next we need to adjust the front end environment variables so close that we're going to open up the front end folder right here we're going to go down to the dot m file now the two environment variables that we need to change are the super base URL and the super base a non-key so let's go grab those from Super Bass and so in that same API settings tab right here it's a non-public so we're just going to grab that and we paste it right there next we need the public URL again so switch back to Super Bass we're going to copy that URL right there switch back to our code and then we input it right there then hit save everything else can stay the same next we have to run this creation script and this is where it got a little bit confusing but it should be simple enough once I walk you through it so there's a little link right here creation script one go ahead and click that and it opens up this creation script and essentially what we're doing here is we're going to feed Super Bass a bunch of commands to create the database the tables that we need and the columns that we need so at the top right we're going to click this little copyraw button switch back to Super Bass and then on the left side there's a SQL editor button go ahead and click that we're going to go to the top left new query new blank query and then we're just going to paste in the contents of that script then right here we click run and if it ran successfully it'll say success no rows returned and that's it now we have all the tables we need to run this all right the last thing we need to do is get Docker up and running so to do that we already have a Docker compose file this is probably the easiest part you just type out Docker compose Dash F and then you reference the yml file Docker Dash compose.yml up Dash Dash build and then you hit enter and then it should build over a few seconds the first time you run this will definitely take longer than the subsequent times and remember you have to have Docker up and running for this to work if you get an error a Docker error it's probably because you don't have Docker up and running on your computer okay it's done and we can see right here the URL that we're going to access this at is localhost colon 3000 so go ahead and copy that and the server is up and running it's waiting for our commands let's make sure it's working switch back to the browser open up a new tab localhost 3000 hit enter and there it is now it looks like it's already logging me in because I already have one up and running but the first thing you'll see is actually a login screen so I'll log out and I'll show that to you now so the first time you get this up and running this is what you're going to see and again this is a pretty full featured application you can put this on production and have people log in to chat with all of your files privately publicly anything you want now if you're signing up for the first time you can do the login with Google if you filled out that information in the dot m file but since we didn't this is not going to work so what we're going to do is don't have an account sign up and then right here you put in an email and password and it will actually send you an email to verify you get that email you click verify and then you should be good to go so I'm going to sign up with my email address and then click sign up and then it says confirmation email sent please check your email I'm going to go do that now hit enter and there we go I'm logged in so now I can log in with that username and password going forward and the first thing we're going to need to do is create a brain so up here with this little brain icon we click it click a new brain I'm just going to call it test and we click plus and there it created a new brain click on that and it should be good now I grab any document and I drag it right into there upload and there now it's chunking it it's creating and embedding from it and it's uploading it to the database all right there it is file uploaded successfully so now I have everything in here and if I go to the explore tab I can see that the document is there and if I go to the chat tab I can now chat with that document so what is this document about there it is the document you provided is the Constitution of the United States it works you can drop anything you want in there it is so cool you also can actually verbally chat with your documents with this little button right here you can actually talk into your mic and ask questions and then they have the settings button right here and this is where you can select your model you can adjust the temperature adjust the tokens and a few other options it also has chat history which is really nice and again probably the coolest thing about this is you can put this on the internet and have it walled behind authorization so you can give access to your friends your colleagues to chat with documents that you create so that's it I think this is incredible I hope you enjoyed this if you like this video please consider giving a like And subscribe and I'll see in the next one

Channel: Matthew Berman

Views: 91,839

Keywords: chat with docs, chatgpt, ai, openai, open source llm, chat with docs gpt, artificial intelligence, privategpt, private gpt

Id: rFEbz93G9U8

Length: 11min 1sec (661 seconds)

Published: Fri Jul 07 2023