Open WebUI deployment and hardening as the remote chat front end
Parent: Local LLM Inference on Consumer Hardware (RTX 5080 eGPU, Linux, Remote Access) · Topic entry · 15 branches · skill ai-llm-model-layer
A published reference is not available for this topic yet.
Children
- Pinned-tag docker compose bound to loopback behind tailscale serve (frontier)
- Host backend reachability via host-gateway and docker0 binding (frontier)
- First-admin, signup and pending-role behavior (frontier)
- ConfigVar persistent-config trap on environment variables (frontier)
- Per-model presets, global defaults and the num_ctx override (frontier)
- Native tool-calling default and its context cost (frontier)
- RAG embedding offload and web-loader SSRF controls (frontier)
- Tools and Functions as root-equivalent code execution (frontier)
- Community function supply-chain risk (frontier)
- Advisory history and vendor stance on plugin attack chains (frontier)
- License branding clause with the 50-user exception (frontier)
- Image size and RAM footprint by tag (frontier)
- Upgrade, one-way migrations and cold backup (frontier)
- Tailscale identity-header SSO caveats (frontier)
- Open WebUI versus LibreChat, AnythingLLM and Jan (frontier)
Frontier under this node: Advisory history and vendor stance on plugin attack chains, Community function supply-chain risk, ConfigVar persistent-config trap on environment variables, First-admin, signup and pending-role behavior, Host backend reachability via host-gateway and docker0 binding, Image size and RAM footprint by tag, License branding clause with the 50-user exception, Native tool-calling default and its context cost, Open WebUI versus LibreChat, AnythingLLM and Jan, Per-model presets, global defaults and the num_ctx override, Pinned-tag docker compose bound to loopback behind tailscale serve, RAG embedding offload and web-loader SSRF controls, Tailscale identity-header SSO caveats, Tools and Functions as root-equivalent code execution, Upgrade, one-way migrations and cold backup