Hugging Face Chat UI

HuggingChat at hf.co/chat runs on Chat UI, Hugging Face's open source SvelteKit chat frontend, and the same codebase self-hosts anywhere Node runs. After a major refactor the app talks exclusively to OpenAI-compatible endpoints configured through OPENAI_BASE_URL, which makes it a clean frontend for Hugging Face Inference Providers, Ollama, llama.cpp server, OpenRouter, or any similar API, with available models auto-discovered from the endpoint's models route and the older provider-specific integrations removed. Features include tool calling through MCP servers, multimodal input on vision-capable models, an optional LLM router that picks a route per request via a configurable Omni virtual alias, per-model capability overrides, and chat history persistence. Setup is a git clone, an .env.local with a base URL and API key, and npm run dev; MongoDB is optional in development because an embedded database persists to ./db, and Docker images bundle everything for production. Licensed Apache-2.0 with 10.8k GitHub stars, it is a common choice for teams that want a polished, provider-neutral chat interface over routed models.