FREE Unstract AI Open Source 🤖Convert ANY Unstructured Data To RAG Ready LLM Data API Endpoints

What's going on guys Josh pook here and if you've ever built a rag llm system retrieval augmented generation if you've been using this with llms or AI agents you will know that data processing data loading is one of the key things when it comes to building good rag agents with quality data good insights and one thing that you may one problem you may have faced is with dealing with unstructured data well in today's video I'm going to show you how you

Can leverage a platform purpose built for M powered unstructured data extraction and this is a no code solution we're going to dive into it and show you how you can set it up fully for free open source let's dive right into [Music] it all right guys so I came across unrack today as I was actually working on a video I've been working on a video for the last 5 days now that is related to Rag and AI agent and whatnot which

Caused me to start looking into unrack because of course you know data loading is so important and crucial in regag all the links I go over of course will be in the description down below unrack here is a no code llm platform to launch apis and ETL pipelines to structure unstructured documents okay so as I started playing around this tool and you'll see here in just a second it is uh pretty robust um it's a no code solution and there's a

Lot more that you can do um and you can get pretty creative with this as you'll see once we get into it that is really just has everything out of box so as you can see here this kind of sums up what uh unstr does so step one do prompt Engineering in prompt Studio step two connect inputs and outputs in workflow studio and then step three deploy workflows to integrate seamlessly so unstructured data apis to custom apps unstructured data ETL pipelines

To Data Warehouse or database and then unstructured Q or custom Q&A apps on private data to non-technical business users okay so I actually made a little document right here this will kind of give you guys pretty much everything um you could need in terms of like unrack resources and a little bit of a a setup guide and a bunch of stuff here so I'll leave this down below for you guys for free so you can reference this if you're going to

Be using unrack now a few prerequisites you're going to need before setting this up is git python Docker and open AI API key there's some info here about why this is important and what the setup and process usually is from uh converting unstructured data to you know llm friendly data I'm not going to go over this whole stuff here but like I said the documents down below so you can check that out um but just to quickly skim through some of

It you know just some examples of unstructured data just could be stuff like text documents so emails reports articles social media posts audio and video files images handwritten notes really anything like you know tax receip seats whatever the case may be um and then so challenges you may face from unstructured data so lack of predefined format or structure variability in language style and quality uh contextual ambiguities difficulty in organizing and indexing and then scalability issues when dealing with large volumes and

Then the process typically of converting the unstructured data is data processing so cleaning standardizing formats removing noise then text extraction so converting non-text data so PDF images into machine readable text and then tokenization and then embedding and then indexing and then the benefits of converting unstructured data so enhanced information retrieval improved decision- making automation of knowledge work uh personalization and customer experience and then Innovation and research so in their GitHub setup they recommend 8 GB of RAM so next thing is

Either download or release or clone this repo to uh and do the following so you're going to want to run get clone and then the repo right here and then you're going to want to CD into unra so um I actually already did that but you're going to open up your terminal and you are going to do that and then you're just going to run this bash Command right here which is uh run- platform.sh so this is once you're in unrack

Now you may want to run uh you may want to actually run code dot if you you don't have have to necessarily but you may want to if you want to check it out some of the files for whatever reason like here just I mean either way it's good to have it open for example like um it will show you what the uh you know Mino uh the minio um username and password is as well as the cint and the uh

Postgress and stuff like that and you can just check out the the whole setup there if you want but it's really you don't need to so yeah once you run this it should start okay so you're going to have to have Docker um open up before you run this so make sure you have that open and then it's going to start building out the containers once you do that you can go to front end. unrack do localhost in your browser and

That's going to take you so it's actually going to take you to a login page um initially and you're going to want to click on log in and then you're going to want to log in with unrack as the username and then unrack as well as the password okay then you're going to log in and you can see user guide right here for more details on managing the platform and if you are having trouble or maybe you don't have enough RAM

You could sign up for the hosted version I think there is a free trial too so personally I personally wouldn't pay for it so yeah I just wouldn't pay for it because I think it's pretty pricey I would rather just leverage the selfhosted version um but yeah you can check it out try it out if you want on the hosted version and then setup so once you go to the you know the set the login here and you log in then

You're going to want to do these four steps so you're going to want to connect an llm and then you're going to want to connect a vector database and then you're going to want to connect an embedding model and then an a text extractor okay so it's going to look like this it's like you're onboarding so there's many different llms you can use you can use olama which is open source models which is pretty cool um I'm just going to use

Open AI for this example and then enter name of llm so you're going to get this um Modo popup right here so you're going to want to enter this name right here you can enter anything it doesn't matter it's just it's just for uh unrack naming uh convention so whatever you want to name it and then you're going to want to enter your API key if you're using open Ai and then the model name so this one you're actually going to

Want to enter the exact model name of whatever model you're using so if you're using open AI you can either just put GPT -40 or if you want to use a different model you can go to the models docs right here find it and then enter it there then you're going to want to click test connection and then click on submit so that's how you add the llm then you're going to want to connect a vector database so you can you

Have four different options you could do um post grass pine cone cudr whatever we actually you have five different options um okay so for the sake of this video we're doing cant because I uh it once you self-host it it automatically sets up a currant database for you um yeah and then the name can be anything just like I said with the llm but I'm just going to use cant and then the URL would be this right here for cant so

It'll be host. doer. internal and then uh 6333 and then the API key can literally be anything since we're self-hosting there's no API key and then click on test connection and then click on submit and then you're going to connect an embedding model so there's four different embedding model options um pal AMA open Ai and then Azure we're going to use open AI here just to keep it simple and you can enter the name again right here this doesn't matter whatever

You want and then uh you're going to want to enter the model name exactly as it is so you can go to embedding models right here on open AI for example I'm going to be using text- embedding -3- small and then click on text connection and then click on submit okay and then four is connect a text extractor so there's four different extractors I'm using llama parse because L whisper is made by unrack I think they give you about a 100

A day but llama parse gives you a th000 free a day so uh if you go here it will ask you to create an account if you don't anyways or log in so you can do that and then grab an API key and then you're going to want to put your API key here you can put any name here again and then you could change the result type to text nextt or markdown whatever you want and then you could toggle on

Verbose or not okay I just left it on in default and then click on test connection and then click on submit are you tired 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 considers

That work 24/7 never get tired and book appointments while you sleep trained on thousands of successful conversations our AI agents El performed human teams at just on10th 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 okay once you do

That you got your four main um quick starting things set up we now we're going to actually set up some stuff and basically show you the basics of doing some extraction of unstructured data um so first things first you're always going to want to start in the prompt studio and I'm going to actually show you um firsthand in just a second like I said there is a lot of different stuff you can do so I kind of want to give you

An overview first so for every new document or class you need to structure you should start by creating pro project studio um project so project studio is a no code prompt engineering environment in which you can load sample docks or select any combination of llm vector DB and Bing model and textractor and see what kind of results you're able to get so this is basically like what you can like do to like plan everything out to ensure that it's working and

Whatnot and then later on if you wanted to you know set up the API you could do so if I go here to prompt Studio you'll see that on the left hand side here there's your setting so if you want to add more llms or vector DBS or embeddings or text attractors you can do so and then there's the build section so you can build your prompt Studio to create structured data from unstructured documents there's the workflow so you can build

No code data workflows for unstructured data there's manage so there's API deployments so unstructured to structured apis ETL pipelines so unstructured to structured data pipelines and then task pipelines ad hoc unstructured data task pipelines okay so there is a lot and there's really like a lot you could do um we won't get to cover everything in this video but I will kind of give you some of the basics and I will give you a road map too if you want to

Expand upon this and go deeper into this okay so when you're in prompt Studio here you can click on new project and this is how you could create a new project here so in this example I'm just doing credit card processing or parser right here uh you can put anything for the author or org name I just put JP and then the description so this project uh parses credit card statements you could add an icon too if you want then click

Save here's just the picture showing you kind of cuz once you actually set everything or get into the prompt uh prompt Studio it's going to look something like this yours won't look like this right off the get-go because um I actually already added some prompts and some documents here but um yeah this is like kind of the different components of the prompt Studio we what you're going to want to do to start is go to the settings up here and you're

Going to want to create an llm profile so in prompt Studio you click on settings and then you're going to click on manage llm profiles and then add a new llm profile okay so so if I go up here to settings you'll see llm profiles I already have one but you could add a new one like so so you can see here I could fill it out like I could do my Vector database text embedding text extractor right llama part you

Could add chunk sizes and overlaps and then you add a name right here and you would just click add okay you can do things like manage grammar um now this Preamble is basically something you want to add as well and the post amble is in too so preambles are prefixes to any prompt instruction that is sent to the llm it can be very useful to avoid any unnecessary rep repetition and then post ambles are like preambles and serve much of the

Same purpose except they are suffix to each prompt you enter so in this case our Preamble is your ability to extract and summarize this context accurately is essential for Effective analysis blah blah blah blah blah you can go ahead and read that if you want and then our post amble is do not include any explanation in the reply only include the extracted information in your reply okay so once you create your llm profile then you can actually start creating prompts so

If I click on new prompt right here you'll see I get a like a new prompt down here now I'm going to delete this because I don't need this but you can see here I already created five different prompts this is like essentially data that we're going to be extracting from here so I loaded up you can click on manage documents here and you can upload files right so I already uploaded some right here once you upload them they'll say not

Indexed um but if you click here you'll see that they actually start indexing and you can go ahead and select whatever document you want to show right here okay so when you're creating a prompt you'll see there's different things that you can change within the prompt creation all right so you can see I got this um dummy document right here as a bang statement loaded up and let's say we wanted to you can see here we got like credit card issuer

Name customer name customer address payment info spend line items and we got you know a description here let's go ahead and say we wanted to run um we could either run all llms for current document or run all llms for all documents let's just do this and then this all right so you can see here we're extracting everything we're extracting the customer name which is Joseph pson right here it's already done now we're extracting the customer address and you can see

The logs down here too um you can see how it's running extracting the different document uh the the different uh line items right here so we can see the results right here we can see that in our prompt we said put all these items into a Json array and return that so we got that right here here as you can see all of the documents have an index now and also too you can either view the document right here in PDF

View or in raw view now you could Click on each one of these like uh separately like I just did or you can go up here and click on them up here to run every single one and then right here is the output analyzer so if I click right here you'll see we can see um you know our document on the left hand side and then we can see the extraction right here on the right right and then you can click

On document parser or on combined output to see everything all right so once you actually you know prove your prompt Studio you can come up here so you can click export as tool here going to Custom Tool exported successfully and now we can use it in different workflows so there's really two main workflows one is API workflows the next is um unstructured data ETL pipelines so API workflows basically allow the publishing of an API that will return Json format when a

Document is passed to it and the ETL pipelines essentially you can pick documents from a network Cloud file system Object Store you can structure them and then push that data into a database or a data warehouse okay so so when you're in workflows you can click on create new workflow then you can add your workflow name right here add a description then click on create workflow that's going to open up the workflow studio right here so first things first is the

Input configuration up here so you could do API or you could do file system we're going to do API but if you wanted to use a ETL pipeline you would do file system and then you can next can connect things like Google Drive Google storage Dropbox uh the minio 3 uh S3 let's go ahead and click API and we're going to do API right here for the output as well all right so on this left right here we have the workflow

Chain so here you can basically drag and drop different tools so I can drag our credit card statement par uh parser right here we also have the default ones such as file classifier text extractor but all your tools are going to show up on the right hand side and you can kind of create more advanced workflows and have step one step two step three um but we're just going to use a basic one right now okay if you click on the

Tool right here you'll be able to access tool setting and then down here we have some other options we have like run workflow clear cache clear process file history and then you can deploy as API and then here is deploy as ETL and then here is deploy as task pipeline so we're going to do deploy as API and then here you can add API deployment okay so you can add your display name description and then an API name and we're going

To click save so if we go up here to API deployments this is where we will see our API deployment so here we can see credit card statements API and we can see our API endpoint so if we go here to manage Keys we can see our API keys right here all right guys so I think you get the point there is a lot more you can do with ract but I just wanted to give you a quick overview on a

Few different things from building workflows prompt Studio API deployments and all the different settings with an LM Vector DBS embedding text extractor Etc so if you're new to the channel we upload videos every single day on AI automation business growth Etc other than that guys thank you so much for the recent 8K subscribers we are on the road to 10K I know we're going to hit it very soon so I appreciate all the new support and uh hope you enjoy this

Video let me know in the thought comments down below if you're going to use unrack Maybe you have tried it before maybe you see a specific use case or maybe you think there's better Alternatives maybe you want to custom code your own thing or maybe you uh just think you have a better tool whatever the case may be I think this is a good tool really um for doing things at scale with the different API deployments and the ETL pipelines are

Pretty cool and I think there's a lot more like I said that you can look into and get some cool use cases from that I didn't even Explore yet so if you have some let me know in the comments down below other than that guys if you haven't already joined our free community stri community.com check the link below you can join our free Facebook group and Discord Channel also too guys if you run a business or you run a call center

And you want to have ai appointment Setters booking appointments into your calendar automatically while you sleep through Voice through text through email whatever the case may be then check out St agents.com I'll leave a link down below speed to lead is super key when calling your leads so our AI appointments setting agents are trained on thousands of successful conversational Pathways to make sure that no matter what industry you're in no matter what business you are in our AI agents will out

Perform your current human appointment centers or call center reps guaranteed so check that out in the description down below and I will see you in tomorrow's video guys keep hustling keep grinding and of course accelerate your stride take care