789 - Tu propio Laboratorio de IA (adios a las subscripciones)

In this episode, I welcome you to an exciting new series where we'll democratize Artificial Intelligence. For the past few months, I've been hearing from great friends and colleagues about the wonders of AI, but the same question kept nagging at me: why do we have to delegate all our information to the cloud? If you know me, you know I'm passionate about **digital sovereignty**. That's why, after concluding our adventure with Podman, I've decided to dive headfirst into the depths of Language Models (LLMs) so you too can have your own "digital brain" running on your own machine. Without depending on third parties, without ads, and, most importantly, with total control over your data. 🧠 Why Local AI? It seems that these days, if you're not using Gemini or ChatGPT, you're out of touch, but the reality is that working in the cloud has its risks. In this video, I'll explain the three compelling reasons to bring AI home: 1. **Privacy**: What happens on your computer, stays on your computer. No more sending private documents to unknown servers. 2. **Cost**: Intelligent agents can be a money pit if you leave them running loose in the cloud. Locally, the cost is zero (beyond the electricity your graphics card uses). 3. **Latency**: While cloud-based models are powerful, nothing beats the immediacy of having the service just a click away on your local network. 🛠️ The Technical Stack: The Rich Container As you might expect, we're going all in here. We're not going to install things willy-nilly and clutter our operating system. We're going to use *Podman* to keep everything organized and secure. I'll explain why I prefer Podman over Docker: basically because of the security of *rootless* mode (without administrator privileges) and the absence of daemons running in the background. Also, I'll introduce you to **Quadlets**. If you missed the previous series, don't worry, here I'll tell you how to turn your AI containers into "first-class citizens" within Linux, managed directly by Systemd. It's truly amazing! 🔧 Tools to get the most out of the lab In this episode, we don't just talk about theory. I'll introduce you to *Ollama**, the heart of our lab, and how I deployed it using a tool I programmed myself to make our lives easier: **QCTL* (Quadlet Control). With a single command, we get the whole setup up and running. We'll also take a look at how to monitor our NVIDIA card with **NVTOP**, because seeing how our GPU works is part of the fun of tinkering. 📅 A 32-Episode Plan This isn't a quick, useless 5-minute tutorial. My goal is to guide you through the next four months (32 episodes) on a journey from zero to one hundred. We'll talk about RAG (Recovery Augmented Generation), agents, skills, MCP, and how to create a true digital repository with all your Markdown documents. Get ready, because we're going to have a blast setting up our own smart environment. Let's get tinkering! --- 📑 Episode Content: 00:00:00 Introduction and the End of the Podman Era 00:01:21 The Slimbook Push and the Linux Center 00:02:15 The Problem of Relying Exclusively on the Cloud 00:03:15 The Master Plan: 32 Episodes of Practical AI 00:05:33 Three Reasons for Local AI: Privacy, Cost, and Latency 00:07:25 The "One-Stop Shop" Philosophy: Squeezing Models with Scripts 00:08:08 The Technical Stack: Why Podman and Not Docker? 00:09:40 Advantages of Rootless and Security in AI 00:10:59 Quadlets: Full Integration with Systemd 00:11:53 Tools: Fish shell, Rust, and Go at the Service of AI 00:13:20 Creating Our Own Digital Memory (RAG) 00:14:00 Directory Structure and Git Repository 00:15:37 The Symbolic Link Trick for Quadlets 00:16:02 Hardware: NVIDIA and Leveraging the GPU 00:16:40 Deconstructing the Ollama Container 00:17:54 QCTL: My Tool for Easily Managing Quadlets 00:20:20 Checking that Ollama is Responding (CURL and API) 00:21:15 Monitoring with NVTOP and VTOP 00:22:13 Farewell and Next Steps in the Local Lab #AI #Linux #OpenSource #Ollama #Podman #SelfHosting #ArtificialIntelligence #DigitalGadget More information, links, and notes at https://atareao.es/podcast/789 🌐 Find everything here 👉 https://atareao.es ✈️ Telegram (the group) 👉 https://t.me/atareao_con_linux ✈️ Telegram (the channel) 👉 https://t.me/canal_atareao 🦣 Mastodon 👉 https://mastodon.social/@atareao 🐦 Twitter 👉   / atareao   🐙 GitHub 👉 https://github.com/atareao