FREE: Ragflow Open-Source GUI AI Agent Rag Graph!🤖 Vector & Semantic Search Self-Hosted 🔍

All right what's going on everybody Josh pook here and in today's video we're going to be looking at installing rlow so rlow is an open-source retrieval augmented generation engine based on deep document understanding it offers a streamlined rag workflow for your business of any scale combining llm models to provide truthful question answering capabilities backed by well-founded citations from various complex formatted data all right we're going to right into it show you how to install it get it set up and running

On your machine all open source and of course all free let's dive right into [Music] it all right guys so here is the raglow website link to the GitHub link to the website and the docks will all be in the description down below just to quickly go over what is raglow as well as some of the features of raglow as well as why is rag flow different than a lot of the other tools we covered on here so rag flow build

Generative AI into your business retrieval augmented generation engine to unleash your full potential so I'm not going to go into the depths of you know what uh rag is but just to somewhat summarize um rag is essentially a way that uh you can get higher quality output and querying and and whatnot of your own data so you can chunk your own you basically do uh there's a multiple different steps but uh it goes through a chunking process as well as a

Embedding process which basically takes a text that you give it let's say you train it on a bunch of knowledge maybe from your company maybe like a bunch of PDFs or whatever the case may be this is as an example um it's going to take that information that data chunk it down into uh different size chunks and use an embedding mold to embed it into somewhat like of a a different code and in different dimensions and then when the user queries

Say if I ask a question trying to get information ask it about my knowledge base that we just put into a vector store into a uh with the embeddings um then it would take my query put it in an embedding that says well a dimensional kind of code um like a um a vector and uh then it would basically use mathematical calculations like cosine or similarity or dot product to uh determine what's the most relevant um you know answer to that

Query and then give me the ideal information from my um data source from my database or whatever the case may be right so there's different types of ways to do rag and everything like this raglow is essentially a tool I'll just quickly go over some of the uh features here so rounds and citations for reduced hallucinations quick view of key reference and traceable citations to support grounded answers so rlow is essentially supposed to be like a tool to to put your

Rag here on on steroids essentially or just give it more capabilities and as well do this in a very uh userfriendly easy way you don't have to like code things or do any of that it's like pretty it's all no code through a gooey and everything like that so we'll we'll show you in just a second quality in quality out so deep document understanding based knowledge extraction from unstructured data with complicated format so that's another thing that rlow is really good

At it's good at taking complicated uh formats I think it's going to show we'll get into seeing some of the formats in just a second but like you could do PDF you could do um you could do uh what's it called um PowerPoint you could do image you could do and I even think they just came out with a new feature for audio so we'll we'll look at some of these uh uh bullet points in more depth to see exactly what

They have on the on the GitHub repo um template based chunking so it's not just like you can chunk it in one specific way they have different templates that you can use for chunking um your information into different ways intelligent and explainable and then compatibility with uh heterogeneous uh data sources supports words uh words slides Excel text images scan copies structure data web pages and more so you can see boom PDF chunk number chunk method General enabled parsing status you know

Uh jpeg dog PT PPT CSV a bunch of different types of data formats all right so that's the website um and they also somewhere they have like a demo you can try out here's the docs I'll leave it in the description down below um but I'm just going to read the docs on at least to start off on the GitHub because I believe it's the same um there may be a little bit more in the Docks but anyways so we here

Yeah here's the demo demo R flow.io and here's some images of kind of what it looks like so they got a new UI Builder which they had it as graph and now they changed it to believe called agents um so supports gra oh wow wow so this is really cool and I'm going to do more videos on um this in the future so supports graph rag um if you don't know graph rag um I'm not going to go too too in

Depth but here's just their uh repo right here it's actually by Microsoft so it's a different way of doing rag graph Rags so um so yeah I'm going to be doing more videos about Rags and different types of rags and not just like the basic rag but like getting even in better high quality Rags show you how to do that stuff um so if you want any specific videos around those topics let me know but anyways um latest update supports audio

Audio file parsing supports llm such as local AI open router um Step fun Nvidia adds more components Wikipedia PubMed all these other ones Duck Duck Go to the graph reports workflow based on graph reports markdown and Doc X in Q&A and parsing method supports extracting images from Doc x file supports extracting tables from markdown files supports selfrag which is uh enabled by default and dialogue settings integrates with BCE BGE reranking models or ranker models supports Raptor for uh better text retrieval

And integrates with open AI gbt 40 so those are some of the key features we already went over and the system architecture is right here so this is kind of like a basic understanding of rag so as we see here documents go in so file goes in task dispatch um into here chunk uh document lay oops let's see here uh document layout and analyze table structure recognition F executor document parser so it uses these different tools these are like different tools

That it will use OCR um so and then chunk and then multi-ray recall rerank uh answer llms so this is when the question the query happens so it goes to the web Eng Gen X question period analyzed so um yeah this essentially just supposed to be a document um showing their system architecture for how the rag works and so prerequisites for using um uh um for using uh raglow is four core CPU 16 GB Ram yeah I won't lie like even

In uh doing the docker comos up it's a pretty like it's a bigger one it's a definitely a bigger one so uh yeah if you don't have the necessary requirements then I mean you could maybe try it but honestly it would probably just be better to do a Docker compose on a server so you can check out like you can get any server header server you can get one uh contabo contabo VPS very cheap like 5 10 bucks put it on

There and it will be good to go and either way if you've been uh following this channel for a while you probably either have a server or like you definitely should get one if you're doing all like if you're doing all these just on your home like desktop or laptop then I would definitely suggest getting a server pay like 5 10 bucks a month and uh you can pretty much do like most of the uh installs that we've done on this

Channel anyways start up okay so and then uh RAM so disk uh 50 gab of space and then Docker so you're going to need Docker and go to doer. download it for your um operating system docker's essential it's it's awesome we use it all the time it's like one of the best things ever so definitely get that regardless and then uh this is how to basically install it so I honestly I just yeah skipped this part I just went to cloning

The repo so I did get clone uh I clone this into my uh repos app on my desktop and then build the pre Docker images and start up the server so it's actually very simple to install and start up it just for it did take me a little bit of time just because it's like a bigger download um but it's very easy to install it just may take a bit of time I don't know if it's going to take everyone as

Long as it took like it maybe took like 10 15 20 minutes but I was also doing other stuff too but anyway so get clone so you would just open up your terminal and if ever like um you're encountering issues with the terminal sometimes you got to run it in as a as administrator but and I don't even think you have to for this so yeah so you would just go get clone the repo which I already have it and then

You're going to want to CD into so this is change your directory into raglow SL doer so once you um get clone it you just run this command okay I didn't run this this is like a Linux command but um yeah if you need to run that on a Linux or whatnot you could I just did right after that Docker compose up it's dasd okay so that's why you need do ER and then if you're run running into any issues you

Could check the server logs for raglow but uh it will um and also too guys you you need to have Docker desktop opened um on your computer so just once you install it make sure it's open so you're good to go on that oops come on oh goodness why does that always happen okay here we go rag flow okay so once you're done that it's going to be open on Port 80 of your server on your your uh Local Host right

If you're doing on your Local Host so yeah in your browser enter the IP address of your server and log into raglow and it's on a uh Port 80 so if you're doing it on a server you put your IP address if you're doing it on your Local Host you can just put Local Host and then um ins servicec config.yml you can select the desired llm Factory in your in user default llm and update the API key with the corresponding uh

Key so see llm API key setup for more information and there's also different configurations you can do um let me just see here open oops go open okay so um yeah this is all about configuring your API keys and uh and whatnot and there's you'll see there's a bunch of different Integrations we're going to get into that in just a second so configurations when it comes to system configurations you will need to manage the following files uh .v this is for

Your msql password Mino password um and all this stuff you must ensure that changes to the EnV are in line with what are in the service config file um and yeah there's other configs that you can do and you can also yeah there's other ways you can install it but anyways let's just get like I just show you the basics of how to get it started from that point on if you've done everything you should basically have your um rag flow

Set up so let me just go back to the homepage here so Local Host all right I'm just doing it on my Local Host right now um I I'm May uh set it up on my kify server uh later so if you do want me to like show you specifically how to do that let me know um I'm also doing some extra some more videos too on showing you guys how to set specific stuff on up on coolify so not familiar

With kify I've done like 10 15 videos on it maybe by now so uh definitely something to look into and if you know about my if you've been following this Channel and you know about like the the perplexity ones I've been doing I have a video coming doing a comparison between different perplexity Alternatives and I'm also going to show you how to install them on coolify in the same video so make sure to stay up to date for that anyways um

So just to quickly go over kind of what we're looking at here we got our profile which is basic you can change like your username your photo your time zone all that good stuff password pretty basic Modo providers is where we get into the good stuff so we got added providers interesting okay they got this is like a rlow I believe it's built by like a Chinese team or something um yeah China's going hard in AI right now so uh yeah

And as you can see they got at it looks like this default um like quen right here is at the top this is a this if you're not familiar with quen definitely really good model especially for quen quen V uh quen 2 for coding um and it's made by a Chinese team uh so show more models we can see different models here embedding models so you would want to put in in API key there can add model so add open AI

Key the base URL so um this this is actually what the base URL is for open AI um right here and then you can put your API key here um we got moonshot you got a bunch right you got olama right base URL um this you could click here okay it's not showing you um okay I'll quickly try to to explain how you would do a llama just cuz I know some people may be asking that uh you would just put

Your base URL here which depending it's it's either going to be local host or um like I okay I would just honestly ask perplexity like what's the base URL for a llama or ask like chpt and it will tell you it's it's like either local host or Docker I forget the docker one but it's that and then the port is I forget but you could get the port just ask GPT you could put that there and then you put your model

Name you can do embedding rerank image IM uh image to text lot of different options we have right and um yeah I'm not going to go through them all Bedrock mistol voi LM Studio Gina and you can see what each one has so some only have llm some have llm text embedding uh image to text got open router Nvidia step to fun or step fun um really a bunch of different ones um you got a lot of different options which is

Pretty cool System model settings so we can see quen plus quen turbo whatever model you got embeddings right here okay your image to text model your sequence to text model rerank model all right um I'll definitely be connecting more of these you got system here so elastic search active primary charts um okay so this is just showing some B info about the system and everything that's set up on uh Team so can set see your team workspace can't really do much

Here it looks like it says an upgrade button but it's not letting me click it um so okay let's get into the main bread and or the main meat and potatoes we got knowledge base chat agent and file management Okay so so knowledge base we can create a knowledge base I just actually here let me create a new one so test okay from this point on we got configuration so you can add a knowledge based photo you can put a description

You can select language which it only has English or Chinese um you can select permissions for only yourself or your team you can select your embedding model your chunk method so as you'll see here as we I'm not going to go through all the different chunk methods but we got General Q&A res uh resume manual table paper book laws presentation picture one audio Knowledge Graph so this will display a visual explanation of the knowledge based categories this is actually really cool

Knowledge graphs uh graph drag and all that that's like I said something I'm going to be doing more videos on um so but it gives a lot of them an explanation of how the chunk method works so supported formats doc X Excel pppt image PDF text um and you can see kind of description and then some images here step one step two so split into pieces merge into a chunk okay um you can select the to a chunk token number right

Let's just put 128 and then layout recognition so use visual models for layout analysis to better identify document structure find where the title text blocks images and tables are without uh the this feature only the plain text of the PDF can be obtained so this is like some cool stuff about rag flow like you may be wondering like okay Josh you showed us different Rags before and like all these different Rags what's different with rag flow so I'd say the thing

That's like rag flow it specializes just with Rags mainly and it that's it has a lot of different options for Rags um in a very easy to use Ma manner and and like a gooey and it can work really well with different formats all right so like like I said like all those different formats whether it's Excel images this that whatever that's why it's uh good so use Raptor to enhance retrieval recursive abstractive processing for tree organized retrieval referred to hugging

Face papers here so you can turn that on if wanted but uh once you're done there you can go retrieval testing so you can see like um final step after success leave the rest for um else basically similarity threshold so I'm not going to but yeah similarity keyword similarity weight you can change that rerank model you can select top K all right you can just hover over the question mark and we'll explain what it is if needed and you can go

Test te uh test text okay and then your data set so you can add a file or create an empty file like if I went here let me just select a file real quick on my desktop um we can just go yeah I'm just uploading this so this is like a vs code thing and then you can click this and it will start parsing so you can see the status there all right so you can see parsing success here so you

Can see if you hover our parsing we can see process started process duration 48 seconds process message Tas has been received OCR is running OCR finished layout analyze finished analysis finished table analysis finished text merging finished finish slicing files start to embedding the context and then finish embedding start to build index and done so really cool you can see the chunk method here you can rename delete download so let's go ahead and just do a bit of testing here so um

Okay so actually let's go to chat create an assistant test so you could do assistant settings upload assistant opener um show quote self rag uh knowledge base so we're going use our test knowledge base go to prompt engine you can see your prompt here we'll just keep it basic for now but you can change that you can select a rerank model right you can go model settings so we can select your model you can change all these different settings and let's

Just keep it very basic for now okay and then if you came over to your assistant you could go ahead and start talking to the assistant all right then if you went over to your assistant assistant you can go ahead and start chatting with the assistant also here we have file management so you can also see all your files within the knowledge base and there we have it and you can add new files here or add new folders here under agent

Is where we can actually create agents so you can either start with blank or nothing or you can create from templates such as web search assistant a chat assistant that combines information both from knowledge base as well as web search engin it integrates with information from knowledge base as relevant search engines to answer a given question what you need to do is setting up the knowledge base in uh what you need to do is setting up the knowledge base in retrieval

Chatbot uh HR callout assistant Chinese uh customer service and interpreter and interpreter um yeah so let's just go ahead and click on this one the web search assistant as a test call this test T just so you can kind of see the interface here so here is kind of what it looks like begin interface get keywords duck ducko B do wikkipedia llm search knowledge base um you can see there's different nodes here such as retrieval generate answer categorize message relevant rewrite

Keyword Duo B Wikipedia PubMed AR archive and Google and Bing so not a huge amount of options like we've seen in some tools like you know Lang flow and you know Diffy or um you know uh flow wise or what whatever the case may be we done Lots videos on this but this is like I said this tool strength um is mainly around if you want to try out rag in different ways with different types of file formats and you want

To do it an easy easy deoy way um that's when I would look at rag flow and I still think it's I think it's a really good tool and a really useful tool in certain cases um and yeah I'm going to play around with it more and see what I can do with it but if you have any questions about getting it set up or you know anything we covered in this video let me know in the comments down below you

Want to uh make any specific videos on other topics um that you know you want me to cover let me know in the comments Down Below guys any the channel we upload videos on AI automation business growth Marketing sales anything around that type of stuff we upload videos like that every single day so if you're new to the channel and you got some value here make sure to smash that like button smash up that subscribe button to stay up to date

With the daily uploads I got a lot of cool cool value stuff coming for you this is just not even the tip of the iceberg so um we're just ramping up here on this channel and also too guys if you're new here and you haven't already join our free community stride Community Link for that will be in the description down below our free Discord Channel and our free Facebook group with a lot of free resources as well you can network with

Like-minded individuals AI enthusiasts entrepreneurs myself Etc we have about 750 800 members and both those combined also too guys youone checked out my book client acquisition code and you want B uh clients for your business check that out very it's like my most cost effective thing it's like a cost of uh book that I worked it's like worked me I worked it it took me multiple years to to finish that book and actually created it was not like a very easy

Task but um yeah so it it has a whole chapter on AI and prompting in there and uh yeah you can learn a bunch of stuff just for barely anything and uh I'm going to be adding more stuff to people who uh features for people who buy that book such as of course content Community as well and all that good stuff so um check that out Link in the description down below other than that guys I will see you in tomorrow's

Video keep hustling keep grinding and of course guys accelerate your stride take care