Live demo

The real tools, demo-locked

This is the same UI I use daily — a self-hosted AI studio running on my own GPUs behind a Cloudflare tunnel. Everything below is live: the worker pool, the pipeline forms (auto-generated by introspecting ComfyUI workflow graphs), the websocket progress bar, and a local LLM. Demo accounts get preset prompts and one run per day; sign-in unlocks all of it.

Image Studio

Browse every pipeline to see its auto-generated form. The text-to-image pipeline is runnable in the demo.

Loading pipelines…

LLM chat

The same chat interface as the private app, answered by Google's Gemma running locally via Ollama on my GPU — no cloud API.

Pick a prompt above — the reply streams live from a local model (Gemma via Ollama) running on my GPU.

Curious how it's built? The stack: Next.js 16 + TypeScript, a worker pool dispatching to two ComfyUI GPUs (RTX 3080 Ti + GTX 1080 Ti), automatic workflow introspection, Auth.js, PM2, and a Cloudflare tunnel — all covered on the home page.