Open WebUI at Home: The Defaults That Bite Later

Open WebUI is the usual way to give a household a chat page for its local models: one container, users, documents, web search, and an interface much like ChatGPT. It is also easy to deploy in a way that bites later. The bigiron.cc guide deploys it for a household or small team, with defaults read from the v0.11.4 source (released September 2026). Leave the secret key unset and everyone can be logged out after an ordinary update. The first person to reach the sign-up page becomes the administrator. Many environment variables stop having any effect after the first start. Admins can read everyone's chats by default. And document search hands the model three chunks of about 1,000 characters, found by a small English-only embedding model. This video walks through each one and what to set instead. If you are still choosing between front ends, see our LibreChat vs AnythingLLM vs Open WebUI video first. What's covered: • What you are deploying: one container, one data directory • A Compose file with a pinned tag, bound to localhost • The secret key that changes by itself, and how to pin it • The first-admin race, sign-up defaults and pending users • Persistent config: why changing an env var does nothing • One or several Ollama servers, and the localhost trap • OpenAI-compatible backends: llama.cpp, llama-swap, vLLM, LiteLLM • Model access control, named models, and the backend context window • Admin chat access is on by default: turn it off at home • Community sharing, update checks and offline mode • RAG defaults: top-K 3, 1,000-character chunks, no hybrid, no reranker • What to change, in order, and why you must re-index • Web search through SearXNG (and the HTTP 403 fix) • Reverse proxy with WebSockets, and remote access over a VPN • Backups, one-way migrations, and a safe upgrade routine • Checks to run before you hand anyone the address ⏱️ Chapters 0:00 Open WebUI at Home 1:05 A Deployment That Lasts 2:41 Accounts and Settings 4:14 Connecting Models 5:51 Chat Privacy at Home 6:50 Fixing Document Chat 8:35 Search, Proxy, Backups 📖 The full guide (the complete Compose file, the backup script, the upgrade routine, verification commands and a troubleshooting table): → Open WebUI on a Home Server: Users, Models, RAG, and the Settings That Matter: https://www.bigiron.cc/guides/open-we... 🛠️ Gear (affiliate — same price to you, supports the channel): • Mini PC for an always-on Open WebUI box → https://www.amazon.com/s?k=mini+pc+32... • 12 GB-class GPU for the Ollama host → https://www.amazon.com/s?k=rtx+3060+1... • External SSD for the data-directory backups → https://www.amazon.com/s?k=external+s... As an Amazon Associate we earn from qualifying purchases. Affiliate links — same price to you, supports the channel. 👉 Subscribe for self-hosted AI + homelab, with new videos every day. #OpenWebUI #Ollama #LocalAI #SelfHosted #Homelab #RAG #DockerCompose #SearXNG #Privacy #LLM