Docker, Node, and Electron Walked Into My Terminal. So I Built a 3.5MB App to Kick Them All Out.

2 AM, 4 GB of Docker, and a Very Simple Question I just wanted to ask Llama 3.2 about a regex. That's it. One prompt. One answer. 30 seconds of work. Here's what actually happened: Open Docker Desktop Wait for the Open WebUI container to wake up Watch it silently consume 500 MB of RAM Open a browser tab pointed at localhost:3000 Finally type the question Get my answer Notice the container is still running, eating my battery at 3 AM For a chat window. A chat window for a local model. I closed my laptop, stared at the ceiling, and decided: this is no longer acceptable. If you've tried to put a UI on Ollama on Linux, you've hit this wall. Your options are basically: Tool Installed Size What You Need Reality Open WebUI 500MB+ Docker daemon + container runtime Heavy, always-on background process Lobe Chat / Hollama 100MB+ Node.js + npm install A web server you have to start and stop Msty / Jan 200–400MB Bundled Electron runtime A whole Chromium browser to send text Browser extensions N/A A specific browser Limited, can't do file uploads properly Just use the CLI 0MB Memorize every flag Writing JSON by hand just to send an image None of these feel like software. They feel like workarounds. And the worst part? Ollama itself is fantastic on Linux — fast, lean, runs anywhere. The problem was never the model. It was everything wrapped around it. Meet Ollama Chat — a native Linux desktop app for Ollama that weighs 3.5MB installed, uses ~50MB of RAM, and ships as a proper .deb package. / ollama-chat Ollama Chat A native Linux desktop UI for Ollama — ~3.5MB, no Docker, no Node.js, no nonsense. 🚀 Get Started Now v1.0.0 • ~3.5MB • Linux (amd64) • Free & Open Source 📋 Table of Contents 🤔 Why I built this ⬇ Installation 🎯 Getting Started ✨ Features 🖼 Screenshots 🛠 Tech Stack 🏗 Build from Source 🤝 Contributing 📄 License 🤔 Why I built this If you use Ollama on Linux, you already know the pain. You install Ollama, pull a model, and then... you're stuck in a terminal forever. Want to switch models? Type a command. Want to use a different model mid-chat? Stop, remember the command, switch, start again. Want to upload an image? Hope you know the exact syntax. Want a chat interface like Claude or ChatGPT? Linux users are left behind — Ollama has a beautiful native UI for macOS, but Linux gets nothing. … View on GitHub It's not a web server. It's not a Docker container. It's not an Electron app. It's a real native window with native titlebar controls, sitting in your application menu like any other app you actually use. Layer Tech Why Runtime Tauri 2.0 Rust + system WebKit, no bundled Chromium UI React 18 + TypeScript Type-safe, component-driven Styling Tailwind v4 + Radix UI Accessible, themeable, fast State Zustand No boilerplate, no Redux Build Vite 5 Fast dev, tiny production output Packaging .deb + .AppImage Real Linux install methods Tauri is the architectural decision that makes this whole thing possible. Electron apps ship a full Chromium + Node.js runtime — that's why they're 150–300MB. Tauri uses the system's existing WebKit renderer plus a thin Rust backend. Result: a real native app, 1/50th the size. I'm not going to give you a feature checklist. Here's what I use every day: Real-time streaming — token by token, the way LLMs are meant to feel Smart model picker with tier grouping (Frontier 70B+, Balanced 7–34B, Efficient <7B) and capability badges (vision, tools, thinking) Drag & drop images into the chat — works with LLaVA, Qwen2-VL, LLaMA3.2-Vision System prompt presets — Engineer, Writer, Analyst, Terminal, Concise Pull and delete models from inside the app, with a live progress bar Full model info — parameters, context length, quantization, size Export conversations as Markdown or JSON Light / Dark / System themes \`bash curl -fsSL https://ollama.com/install.sh | sh ollama pull llama3.2 wget https://github.com/itslokeshx/ollama-chat/releases/download/v1.0.0/Ollama.Chat_1.0.0_amd64.deb ollama serve & \ That's it. No Docker, no Node, no web server, no localhost:3000. App Installed RAM Cold Launch Open WebUI (Docker) 500MB+ ~400MB 5–10s Lobe Chat (npm) 100MB+ ~200MB 2–4s Msty (Electron) 250MB+ ~300MB 2–3s Ollama Chat 3.5MB ~50MB <1s 3.5MB. I had to double-check. Then I triple-checked. Then I checked the .deb file size on disk and laughed. It's MIT licensed, open source, and about 200 lines away from working: \bash \ If you want a real .deb: \bash \ ⭐ If this scratches an itch, a star on GitHub means a lot. And if something breaks — open an issue, I'll fix it. I'm building this in the open. Roadmap lives on the repo: cloud model support, conversation sync, plugin system. If you want a feature, open an issue. If you want to build it, open a PR. Now I want to hear from you: what's the most absurd dependency you've installed just to do something simple? I just want to feel less alone in this.