What's happening all? Today I want to talk to you about RAG. RAG is really the easy way to train your LLM agents with all the data you can give it. It's not like fine tuning, it's more like buildin...
Transcript
Hello, what's going on, everybody? It's your boy, the logical dude. Today, we're going to be talking about the rag. We're going to have the talk. But no, it's not about the rag that you may be thinking of. It's not a kitchen rag. It's not that time of the month kind of rag. It is retrieval, augmented generation. That is what we are talking about today. And what is that, dude, you may be asking? Well, we're going to talk about that. So what a rag database is going to do for you if you're running, especially if you're running local LLMs, all right? This is like a huge must if you're trying to run local LLMs, especially if you're running smaller models, like anywhere between a 3B to, you know, I mean, I have a massive rag database that we're building right now for my 35B models, right? So what a RAG database can do for you is basically help you. Make your LLMs as smart as inhumanly possible. Give them the biggest wealth of information to dig through that they possibly can have, right? It can be as big as you want it to be. You can go out and you can search the entire internet and scrape and pull everything, put it all in marked down format, which is essentially what I'm trying to do here. or you can do very specific amounts of information. Like, say, you want your model to be only good at certain things, or you are building your business, and you want it to be really just fully trained on your business data and all of your business information and that kind of thing. Well, that is what a retrieval augmented generation database is all about. So instead of opening up your model's brain and trying to stuff a bunch of information in, you're just giving it a really nice, organized private library for it to go to first, before it reaches out to the internet or anything like that. This allows you to have a whole bunch of information. offline so if anything were to happen you lose connection or anything like that like for me you know i live on a homestead so i have a whole database of information about homesteading survival gardening you know uh animal husbandry and all you know all that kind of stuff you know how to raise cattle and chickens and sheeps and goats and like i've got an entire database of information that i can then if something were to go offline I can use my stuff and say, hey, I need help doing this. What, you know, I'm in my zone. You know, let's figure out, blah, blah, blah, blah, what I need to grow, what I can, you know, what I can do to get some food, or, you know, how do I build shelter, or how do I filter water, how do I do this, you know? Whatever may be, whatever's important to you, that's what you can make your rag database all about, right? So... check into that kind of stuff, especially again. If you're running local LLMs, it's really essential that you build a rag database. I have a tool that I have, well, you know, with the help of Claude. Yeah, I think, yeah, Claude was the one that built it. Yeah, anyway, Claude built the RAG database tool, the MCP tool that I have in my agent software, right? I'm going to work on getting some of my tools kind of generalized, because right now they're very specific to my computer and my information and that kind of thing. But I want to make them more generic and put them up on my GitHub so that you guys can, you know, if you want to go and pull some tools, you can go and pull some tools and set it up, you know. So, That's one thing I'm gonna be working on here pretty soon. But yeah, if you didn't know that, and you hear all this, like, you know, tech talk about fine-tuning and this and that, and, you know, you need to have this quantization and that quantization, you gotta have all these GPUs to be able to train a model. Nah, man, build you a rag database. it's not that difficult. So there's all kinds of tools that lets your agents go out and scrape the internet and get different documentation and turn it into markdown documents that LLMs absolutely love. They love markdown documents. So, that's what I'm having mine do. I had my planner model. you know, like my project planner model come out, and we came up with the full foundation. Like, if you were going to, you know, start off and build a, you know, building an app, what information would you need? And it started that often. Then we just kept adding on from there, just filling information gaps wherever we can. And, like I said at this point, I'm at, like, phase 34, something like that. We're adding on, like, 10 phases at a time. So I just have it go back through and just say, hey, like, you know, where are the gaps at now? you know, where, you know, where can we complete some information, you know, what would you need if we need to go completely offline that you can go and build any application I need you to build for the most part. So that's what it's doing. But yeah, check it out. R-A-G, that's what it's all about. Some rag database makes your thing smarter. So yeah, go out there, build your stuff. Get active with this technology, and if you get confused, your dude's here for you, man. Yeah, just hit me up in the comments, so you can find me on the Discord. I'm around here. So yeah, hope you guys have a great rest of your Monday. I'm done with work for the day, man. I'm going to go finish cutting grass outside because we're getting ready for a fun little partay this weekend at the Dirty South FPV park, so trying to make everything look nice. We've got a bunch of leaves and pines straw everywhere, so I'm going to go get on the mower and go mulch it up. You guys, have a great rest of your day.