AnythingLLM Create Embed Vector Database With Local LLM Easily

this is the easiest way to build an embedded database or factor database on top of llm with your own documents and create a knowledge base with AI today we want to talk about anything llm now this is software that you can run on top of any large language model and you can use various types of documents like PDFs Microsoft doc files and other supported file formats as well as web links to build a knowledge base or embedded files using using anything llm you can also use this to build a factor database for your companies for example if you have an e-commerce store you can create a factor database to use for customer segmentations and gain a better understanding of your audience now let's check out how we can run anything llm with our offline or local llm on our local machine so first we need to download anything llm from here now we have the executable files execute that and install it it's very easy and simple now we come to the main page the first time we need to click get started running and here I'm going to choose ol now you have other AI connection options you can use the open API and LM Studio among other things here I input ama's base URL choose the models and set the token context windows I leave everything as the default settings and then for the next step I choose the embedder for the local documents and use the default one for other custom documents as well as the lens DB the lens DB is for our Factor database connections again I use the local one as the default but you can also use other Factor database providers I just want to set up a simple local database and local llm connections all on one one machine now there's a survey you can input your information here if you want to receive updates about anything llm I'm going to do that as well maybe I'll just say it's for my side hustle education and you can type your comments or skip this survey if you don't want to fill it out now for the first workspace this is like a chatbot area where you want to work right here let's say I have an e-commerce site and I want to chat about my e-commerce documents learning to understand more about my customers so let's make this an e-commerce workspace right once we create this you can start chatting here but then before that we have to upload any documents related to my e-commerce store so let's say I have a list of order lists of my customers that I can upload as files and I can turn the CSV files into an HTM ml file just for easier readability for anything llm and then I go to the settings I choose the mistr llm that I installed in ama and I keep the default settings value for other settings here then in the first thread by default thread here you can click the gear button the setting button again and you can embed the documents into the large language models but before that let's test without embedding so I will ask where are my customers mostly located now these questions are coming from the local llm without embedding my documents so it doesn't have any information as you can see the reply from the the local llm Mistral okay so after that let's run this with embedding my documents this is a list of orders from my e-commerce store as you can see this is already embedded in my workspace it's cached in the embedder and then we can close this once we're finished and then let's ask the same questions again and now as you can see the large language model is able to retrieve information from my embedded documents learn it understand it and give me the answer it provides accurate locations of my customers based on their shipping addresses so let's try a broader question just give me the customers right based on those cities that are located in the United States from all my orders that I uploaded in this demo so let's try another one let's try something that I want to grab from the internet for for example I have a research paper of Goo lumere that talks about video diffusion models and I have two links right so I add these two links and embed them into a new workspace one is from GitHub and the other is from AR sees.org and both of these links are in HTML format so let's save click save and embed now in this one I will ask some questions about this diffusion model so let's say give me the summary of this video's diffusion model in point form the large language model is able to retrieve information from the embedded documents or the links that I provided here ar.com and it shows me the information in point form now let's say I want to write about 1,000 words about this video diffusion model it is also able to do that because it has the information and it summarizes it helping me build articles about this diffusion model based on the information I provided in the embedded embedder so as you can see all this information has references as well indicating where the information is sourced from and this looks very professional and also has backgrounds and different point forms so let's try another example here let's say I just get rid of the GitHub page and I only have the ariv research paper that is different from the e-commerce workspace as you can see here check one more time the e-commerce workspace is getting the orders of my e-commerce store and the Lumia workspace only contains the arxiv link from arxiv.org research paper so you can work in different workspaces embedding different documents and that is really good a very well organized way for us to work on other projects or different projects in a company and understand more about the information that we have provided so that is a very quick demo of anything llm and I hope you guys get inspired and try it out with your own information this is completely working locally so I will see you guys in the next videos have a nice day bye

Channel: AI Business Ideas @ Benji

Views: 4,604

Keywords: artificial intelligence, Anything LLM, embedded database, factor database, knowledge base, large language model, web links, e-commerce, customer segmentation, AI services, LM Studio, local machine, data analysis, decision-making process, research papers, workspaces, document embedding, intelligent responses, AI-powered knowledge management, OpenAI, AI Chatbot, AnythingLLM, anythingllm install, anythingllm ollama

Id: 3o_Hi7SKtJ8

Length: 7min 33sec (453 seconds)

Published: Mon Mar 04 2024