792 - Open WebUI, el ChatGPT PRO que puedes tener GRATIS y 100% local

Have you noticed that the "everything free" era in artificial intelligence is coming to an end? We're seeing how the big players like ChatGPT, Gemini, and Claude are starting to tighten the screws on pricing. And don't get me wrong, these tools are fantastic and make us much more productive, but... at what cost? I'm not just talking about money; I'm talking about something much more valuable: your privacy and your digital sovereignty. In this podcast episode, I'm bringing you the definitive solution for setting up your own AI lab at home, without depending on third-party cloud services and with an interface that rivals commercial ones. Today we're talking about **Open WebUI**. #### From the Terminal to Browser Convenience Until recently, if you wanted to tinker with language models (LLMs) locally, like Llama 3 or Gemma, you had to wrestle with the terminal using tools like Ollama. And hey, I love the terminal, but I admit that for everyday use, we all prefer a user-friendly visual interface, with its buttons, organized chats, and dark mode. Open WebUI is precisely that: the elegant "dress" we put on Ollama. It's an incredibly powerful interface that you can self-host and that gives you total control. Best of all, if you glance at it, it looks just like ChatGPT! #### Extreme Customization: My Chat with Leslie Winkle One of the things I enjoy most about Open WebUI is the ability to create custom "Models." In this episode, I'll tell you how I configured a model with the personality of Leslie Winkle (yes, from The Big Bang Theory). I told it who my archenemy is and how it should respond to me. The result? An AI that calls me "cork brain" and that I'm having a blast with while testing its capabilities. But beyond the joke, this is incredibly useful. You can create a programming expert, a digital sovereignty mentor, or an assistant who knows all your personal documents inside and out. #### Features You'll Love: **User Management**: If you have a powerful server at home, you can grant access to your family or friends with granular permissions. You decide who can download templates or who can only chat. **RAG (Recall Augmented Generation)**: It sounds complex, but it's incredibly simple. It's the ability to upload your own PDFs or documents so the AI ​​can respond based on your information. Your own private knowledge base! **Web Search with SearXNG**: I've integrated my own search engine so the AI ​​can search the internet for fresh information, but through my filter, maintaining privacy and delivering clean results. **Python Support**: You can run code directly, opening up an infinite range of possibilities for automating tasks. #### My Infrastructure: How I Have It Set Up I'll also explain my technological "guts." Don't worry, it's easier than it seems. I use Podman (an alternative to Docker) to run four key containers: 1. **Ollama**: The engine that powers the models. 2. **Open WebUI**: The visible interface. 3. **SearXNG**: For searches. 4. **Traefik**: The conductor that allows me to log in by typing "openwebui.localhost" instead of remembering impossible port numbers. This is pure digital sovereignty. It's knowing that your questions, your files, and your crazy ideas aren't feeding into the training of a large corporation, but rather stay on your computer, under your control. Chapters: 00:00:00 To the Rich Local Model! Introduction 00:00:27 The end of the "free" era in commercial AI 00:01:31 Privacy and budget: The two reasons for localization 00:02:42 Beyond the terminal: Searching for the ideal interface 00:03:47 Introducing Open WebUI: The brain of your lab 00:05:21 What exactly is Open WebUI? 00:06:28 Extreme Customization: My Chat with Leslie Winkle 00:08:11 User Management and Granular Permissions 00:09:48 PWAs, Markdown, and Mathematical Formula Support 00:10:55 Model Builder: Create Your Own Custom Experts 00:12:12 Integrating Python and Advanced Functions 00:13:34 Searching the Web Privately with SearXNG 00:15:13 Cloud Integrations and Vector Databases 00:16:08 A Look at the Admin and Consumption Panel 00:18:24 The Art of Prompting: Folders and System Roles 00:20:38 My Infrastructure: Podman, Traefik, and Containers 00:22:56 Resources, Cheat Sheets, and the GitHub Repository 00:24:12 Farewell and the Usual Suspects Network. More information, links, and notes at https://atareao.es/podcast/792 🌐 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