Hey HN, I've been tinkering with self-hosting LLMs and got frustrated with bloated setups needing a dozen API keys for chat, search, OCR, and more. So I built llumen: a dead-simple, full-stack chat web app that gets you chatting with models out-of-the-box.

Why it might interest you: - <b>Effortless Start</b>: Drop in a single OpenRouter API key—no juggling endpoints. Docker image or binaries ready for Linux/Windows (prebuilts on release). - <b>Blazing Fast & Tiny</b>: Boots in under 1s, <100MB disk. - <b>Smart Features</b>: Markdown with code/math rendering, normal chat mode, web-search integration and WIP agentic/deep-research modes. - <b>Opensource</b>: MPL 2.0 licensed. Screenshots and a quick demo GIF in the README to see it in action.

Quick Docker spin-up: <pre><code>docker run -it --rm -e API_KEY="your-key" -p 80:80 -v $(pwd)/data:/data ghcr.io/pinkfuwa/llumen:latest</code></pre>

Default login: admin / P@88w0rd.