HuggingFace Chat UI — Open-Source AI Chat Interface

Introduction

Chat UI is the open-source SvelteKit application that powers HuggingChat, Hugging Face's public AI assistant. It provides a polished conversational interface supporting multiple models, web search augmentation, tool calling, multimodal inputs, and assistant personas, all deployable on your own infrastructure.

What Chat UI Does

• Provides a production-ready chat interface for any text-generation model
• Supports multiple model backends: TGI, vLLM, Ollama, and OpenAI-compatible APIs
• Enables web search integration to ground responses in current information
• Offers tool use and function calling for models that support it
• Manages multiple conversations with sharing, starring, and assistant presets

Architecture Overview

Chat UI is a SvelteKit application with server-side rendering. It uses MongoDB for conversation persistence, session management, and user accounts. The frontend streams tokens via Server-Sent Events. Model inference is delegated to configurable backends defined in environment variables, with each model having its own prompt template and parameters.

Self-Hosting & Configuration

• Requires Node.js 18+ and a MongoDB instance (local or Atlas)
• Define models in .env.local with endpoint URLs and prompt templates
• Configure authentication via OpenID Connect or run without auth for internal use
• Deploy with Docker using the provided Dockerfile or on any Node.js hosting
• Enable web search by configuring a search API (SearXNG, Bing, or Google)

Key Features

• Multi-model support with per-model system prompts and sampling parameters
• Web search augmentation with source citations in responses
• Image and file upload for multimodal models
• Custom assistant personas with configurable instructions and tools
• Conversation sharing via public links and export to JSON

Comparison with Similar Tools

• Open WebUI — Python-based, Ollama-focused; Chat UI is JavaScript/SvelteKit with broader backend support
• LobeChat — React-based with plugin marketplace; Chat UI is lighter and battle-tested at HuggingChat scale
• LibreChat — multi-provider with file handling; Chat UI emphasizes HF ecosystem integration
• Chatbox — desktop client; Chat UI is a web application for team deployments
• text-generation-webui — Python/Gradio for power users; Chat UI offers a cleaner end-user experience

FAQ

Q: What models work with Chat UI? A: Any model served via TGI, vLLM, Ollama, or an OpenAI-compatible API. Configure the endpoint in .env.local.

Q: Is MongoDB required? A: Yes, for conversation storage and user sessions. A free MongoDB Atlas tier works for small deployments.

Q: Can I add web search to responses? A: Yes, configure a SearXNG instance or Bing/Google API key for RAG-style web search augmentation.

Q: How do I add authentication? A: Set OPENID_CONFIG in your environment to enable OAuth/OIDC login with any provider.

Sources

• https://github.com/huggingface/chat-ui
• https://huggingface.co/docs/chat-ui